WO2017011756A1 - Biomemetic systems - Google Patents

Abstract

Provided herein are systems containing a microbiome that can have at least one genetically modified bacterium having a synthetic gene circuit, where the microbiome can be coupled to a sensor that can be coupled to a computing device that can have a processor capable of executing an application that can cause the processing circuitry to at least receive a signal from the sensor and send a command to a robotic device to execute a function. Also provided herein are methods of controlling a robotic device by growing a microbiome that can have at least one genetically modified bacterium having a synthetic gene circuit, sensing a biological activity of the microbiome, generating a signal in response to the sensed biological activity, and transmitting a command to a robotic device to execute a function based at least in part upon the output voltage.

Description

BIOMEMETIC SYSTEMS CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of and priority to co-pending U.S. Provisional Patent Application No. 62/193,230, filed on July 16, 2015, entitled“IN SILICO MODEL OF BIOMIMETIC ROBOTS AND ENGINEERING LIVING CELLS,” the contents of which is incorporated by reference herein in its entirety. STATEMENT REGARDING FEDERALY SPONSORED RESEARCH OR DEVELOPMENT This invention was made with government support under Grant No. IGERT 0966125 awarded by the National Science Foundation and Grant No. FA9550-13-0108 awarded by the Air Force Office of Scientific Research The government has certain rights in the invention. BACKGROUND Microbiome-host interactions and the effect of the relationship between the two are complex and thus difficult to study using traditional methods. As such, there exists a need for improved systems and devices to improve the understanding of these significant microbiome-host interactions. SUMMARY Provided herein are systems that can have a microbiome, wherein the microbiome can contain a genetically engineered bacterium that can have a synthetic gene circuit responsive to an inducer and wherein the synthetic gene circuit can be configured to generate an optically active protein; a microscope, wherein the microscope can be optically coupled to the microbiome and wherein the microscope can be configured to receive an optical output from the optically active protein and produce an output voltage correlating to the optical output from the microbiome; processing circuitry having a processor and a memory, wherein the processing circuitry can be coupled to the microscope; and an application that can contain machine readable instructions stored in the memory that, when executed by the processor, can cause the processing circuitry to at least: receive the output voltage from the microscope; determine a characteristic of the output voltage; and provide a command to a robotic device to execute a function based at least in part upon the output voltage.

  Also provided herein are systems that can have a microbiome, wherein the microbiome can have a genetically engineered bacterium comprising a synthetic gene circuit responsive to an inducer; a sensor, wherein the sensor can be coupled to the microbiome and wherein the sensor can be configured to produce an output signal corresponding to the biologic activity of the microbiome detected by the sensor; processing circuitry having a processor and a memory, wherein the processing circuitry can be coupled to the sensor; and an application containing machine readable instructions stored in the memory that, when executed by the processor, can cause the processing circuitry to at least: receive the output signal from the sensor; and provide a command to a robotic device to execute a function based at least in part upon the output signal. The sensor can be optically, biologically, chemically, fluidically or electrically coupled to the microbiome. The processing circuitry is electrically, optically, or wirelessly coupled to the sensor. The system can further include a robotic device, wherein the robotic device can be electrically, optically, or wirelessly coupled to the processing circuitry. The system can further include a chemostat, wherein the microbiome can be contained in the chemostat. The system can further include a robotic deivce, wherein the robotic device can be coupled to the chemostat and/or the processing circuitry. The robotic device is electrically, optically, or wirelessly coupled to the chemostat and/or the processing circuitry. The chemostat can be electrically, fluidically, or optically, coupled to the sensor. The chemostat can contain a microfluidic channel, wherein the microbiome can be contained in the microfluidic channel. The chemostat can be a microchemostat. The system can further contain a microscope that can be optically coupled to the microbiome and is electrically, optically, or wirelessly coupled to the sensor. The synthetic gene circuit can be configured to generate an optically active protein and the biologic activity of the microbiome sensed by the sensor can be a wavelength of light produced by the optically active protein. The application, when executed by the processor, can additionally causes the processing circuitry to at least determine a characteristic of the output signal. The output signal can be a voltage, optical signal, chemical signal, biologic signal, audio signal, or electromagnetic signal. The characteristic can be the average output signal over a period of time. The function can be to move the robotic device in a direction over a distance. The inducer can be an environmental inducer or internal inducer.

Also provided herein are methods of controlling a robotic device. The methods of controlling a robotic device can include the steps of growing a microbiome in a chemostat, wherein the microbiome can include a genetically engineered bacterium that can have a synthetic gene circuit responsive to an environmental inducer; sensing a biological activity of the microbiome; generating a signal in response to the sensed biological activity; and transmitting a command to a robotic device to execute a function based at least in part upon to the output voltage. The microbiome can be grown such that substantially all of the  

bacteria of the microbiome are in the exponential growth phase. The synthetic gene circuit can be configured to generate an optically active protein. The biological activity can be light generated by the optically active protein. The signal can be a voltage, optical signal, chemical signal, biologic signal, or electromagnetic signal. BRIEF DESCRIPTION OF THE DRAWINGS Further aspects of the present disclosure will be readily appreciated upon review of the detailed description of its various embodiments, described below, when taken in conjunction with the accompanying drawings.

FIGS. 1A-1C show living cells interfaced with a biomimetic robot as a model system for host-microbiome interactions. (FIG. 1A) A synthetic gene network– also known as an engineered gene circuit. Uploading gene circuit into living bacteria endows cells with a programmable biomolecular network. (FIG.1B) Engineered bacteria and their circuits can be introduced into an organism’s microbiome. The networks of the host and microbiome combine to form a complete gene network. In the absence of the complete host-microbiome network, host behavior is erratic. A programmed microbiome drives new, and potentially rational, host behavior. (FIG. 1C) A robot with a microfluidic chemostat mimics the microbiome niche in an organism. The robot is conceptualized to include a miniature fluorescent microscope, along with the pumps necessary to deliver inducers to the onboard microfluidic chemostat. This microscope allows for modulations in the reporter protein levels to be interpreted by the robot electronically. In the absence of a living microbiome, robotic host behavior can be erratic. A programmed, living microbiome drives new host robotic behavior.

FIGS. 2A-2F show a computational simulation approach for the model system. (FIG. 2A) A basic gene circuit– the lac-inducible gene network– forms the core of all simulated gene network behavior. (FIG. 2B) Green fluorescent protein (GFP, shown as a green dot) from this circuit is conceptualized to be detected by an onboard miniature, epifluorescent microscope (EFM). (FIG. 2C) A computational simulation of microbiome GFP production based upon an analytical model for the circuit in (FIG. 2A). In a built-system, this protein fluorescence signal would be the light detected by the EFM. (FIG. 2D) The conceptualized robot uses onboard electronics to convert the measured light signals into electrical (voltage) signals. (FIG. 2E) Voltage signals meeting specific criteria activate pre-programmed robot motion subroutines. (FIG.2F) The resulting emergent behavior can lead a robot to a carbon fuel depot. Here, robot behavior resulting of a simulation of the circuit in (FIG. 2A) is shown. The robot was programmed with motion subroutines that activate to seek arabinose (orange square) depots following receipt of a lactose (cyan triangles).  

FIGS. 3A-3E show emergent robotic host behavior resulting from a microbiome with a Bistable Memory Circuit. (FIG. 3A) A bistable switch– or balanced genetic toggle switch - was simulated. The gene topology is represented using systems biology network notation. (FIG. 3B) Simulation results for internal inducer concentrations of lactose (cyan) and arabinose (orange). (FIGS. 3C and 3D) Simulation results for internal fluorescent protein reporter concentrations of mCherry (red) and GFP (green) are shown in (FIG. 3D). These are parsed into the EFM electronic output shown in (FIG. 3C). (FIG. 3E) A simulation of resulting robot motion depicts movement at constant velocity through the arena with stops (larger red octagons) to dock at individual carbon depots.

FIGS. 4A-4D show emergent robotic host behavior resulting from a microbiome with an Unstable Memory Circuit. A biased switch– or unbalanced genetic toggle switch - with the topology shown in FIG. 3A was created by increasing the ribosome binding site (RBS) for LacI to be 2.4 times the strength of the RBS for TetR. (FIG. 4A) Simulation results for internal inducer concentrations of lactose (cyan) and arabinose (orange) (FIGS.4B and 4C) Simulation results for internal fluorescent protein reporter concentrations of mCherry (red) and GFP (green) are shown in (FIG. 4C). These are parsed into the EFM electronic output shown in (FIG. 4B). (FIG. 4D) A simulation of resulting robot motion depicts the robot behaving in a manner different from FIG. 3E, with a clear preference for lactose carbon depots. Specifically, the robot briefly seeks arabinose depots after a lactose depot is acquired, however this period is quickly overwhelmed by the biased toggle switch behavior and the robot changes course to seek out a lactose depot.

FIGS. 5A-5D show results from exploration of toggle switch parameter space. This figure presents how changing RBS strengths driving LacI and TetR expression can change the robotic platform’s behavior without altering the genetic topology. (FIG. 5A) The total number of lactose depots acquired by the robot. (FIG. 5B) The total acquired arabinose depots acquired by the robot. (FIG. 5C) The percentage time the robotic platform spends in stall (i.e., EFM = 0). (FIG. 5D) The total time steps of the simulation. These figures demonstrate behavioral bifurcations driven exclusively by RBS strength.

FIGS. 6A-6E show addition of orthogonal operon yields nuanced predation habits. (FIG. 6A) The toggle switch topology modified with an additional, orthogonal operon containing the P_lux-λ promoter driving polycistronic expression of GFP and mCherry was simulated. This promoter is induced by AHL, which the robot is programmed to inject into the living, onboard microbiome when it nears any carbon depot. (FIG. 6B) Simulation results for internal inducer concentrations of lactose (cyan), arabinose (orange), and AHL (yellow). (FIGS. 6C and 6D) Simulation results for internal fluorescent protein reporter concentrations of mCherry (red) and GFP (green) are shown in (FIG. 6D). These are parsed into the EFM electronic output shown in (FIG.6C). Note the addition of EFM values of 2 and -2 indicating  

the robot is moving at two times the base velocity. (FIG. 6E) A simulation of resulting robot motion depicts the robot moving towards a depot, pausing, and then moving at twice the speed when close to the depot. This behavior appears to be qualitatively similar to stalk- pause-strike predation, an identifiable trait in higher level organisms.

FIGS. 7A-7E show distinct behavioral regimes emerge from RBS modification. (FIG. 7A) The gene circuit topology from FIG. 6A was further modified with an additional, orthogonal operon containing the P_lux-λ promoter driving polycistronic expression of GFP, mCherry, and critically, cI, the repressor from λ bacteriophage. In addition to being activated by AHL, this promoter is also repressed by cI, thus the new operon is auto-repressing. Furthermore, the robot is programmed to inject AHL into the living, onboard microbiome when it nears any carbon depot. (FIG. 7B) When the simulated RBS strength for cI (RBScI) is close to 0.0, the robotic platform behaves in the stalk-pause-strike manner described in FIG. 6. (FIG. 7C) With the RBScI value at 0.0007, there is a decrease in the length of the ‘strike’ period of the predation pattern leading to a stalk-pause-strike-pause-stalk behavioral regime. (FIG. 7D) Increasing the RBScI value to close to 0.01 leads to a regime of inactivity whereby the robotic platform is unable to acquire even one carbon depot. (FIG. 7E) Finally, as the RBScI value approaches 1, the system behaves similarly to the initial balanced toggle switch seen in FIG. 3A-3E. These multiple, and strikingly different, host behavioral regimes indicate how certain biochemical networks of the microbiome may have drastic impacts on host behavior.

FIGS.8A-8C show several embodiments of the biomimetic systems provided herein. FIGS. 9A-9C show a robotic platform information flow. This figure shows a visual representation of information flow through the three modules. (FIG. 9A) The synthetically engineered microbiome, programmed with a synthetic gene network. (FIG. 9B) The microchemostat environment with physical microfluidic channel and epifluorescent (EFM) microscope. (FIG.9C) The robotic host translating EFM signal through a microprocessor into robotic behavioral response.

FIGS. 10A-10C show biochemical model basics. This figure shows a visual representation of the biochemical model used for our simulation. (FIG. 10A) The inducer transport through the membrane barrier. (FIG. 10B) The interactions involved with the translation of [mRNA]. This includes internal inducers, a promoter site and repression proteins. An RBS is also seen as being associated with the mRNA. (FIG.10C) An illustration of the translation event creating a protein, relating in [mRNA] to the [Protein] produced. This process is driven by a ribosome.

FIGS. 11A-11F show exploring stochasticity in the simulated gene network. Six simulations were rim exploring how stochasticity in transcription and translation affected the

 

performance of the robot platform. The reporter protein, EFM signal, and inducer concentrations associated with FIGS.11A-11F are presented in Figures 15A-20D.

FIGS. 12A-12E show a balanced toggle switch with randomly occurring carbon depots. (FIG. 12A) The biomimetic robot host was endowed with bacterial cells containing a balanced toggle switch. (FIGS. 12B-12E) Four different simulations were run with randomly placed carbon depots. Each simulation showed the robot alternating between Lactose and Arabinose depots in a bistable manner.

FIGS.13A-13E show an embodiment of a system information flow that demonstrates how variables can be passed between the five different simulation systems including the three modules from FIGS. 9A-9C, the robotic platform (FIG. 13D) and the arena simulation (FIG. 13E). Chemical, position, and voltage parameters are passed from systems, allowing for modularity of engineering. Initial conditions are shown to the left of the figure. The central, greyed, rectangles represent a simulation group. Finally, the arrows exiting the greyed boxes represent the flow of information, whether concentration, EFM signal, or robot location. FIGS. 13A-13E can be viewed as a visual simplification of the Simulink model that can underlie embodiments of the simulation.

FIGS. 14A-14D show stochasticity in the internal inducer concentration. A Gaussian multiplier was applied to the lactose and arabinose internal concentrations. This multiplier had a variance that was 10% the previous time step’s concentration. (FIG. 14A) The EFM signal for the stochastic circuit. (FIG. 14B) The lactose and arabinose internal inducer concentrations, with a callout box demonstrating the stochasticity of the signal. (FIG. 14C) The reporter protein levels. (FIG.14D) The emergent robotic behavior within the arena.

FIGS.15A-15D show stochasticity in a balanced toggle with 0% transcription and 1% translation variance. (FIG. 15A) The EFM signal for the stochastic circuit. (FIG. 15B) The lactose and arabinose internal inducer concentrations. (FIG. 15C) The reporter protein levels. (FIG.15D) The emergent robotic behavior within the arena.

FIGS.16A-16D show stochasticity in a balanced toggle with 1% transcription and 0% translation variance. (FIG. 16A) The EFM signal for the stochastic circuit. (FIG. 16B) The lactose and arabinose internal inducer concentrations. (FIG. 16C) The reporter protein levels. (FIG.16D) The emergent robotic behavior within the arena.

FIGS.17A-17D show stochasticity in a balanced toggle with 1% transcription and 1% translation variance. (FIG. 17A) The EFM signal for the stochastic circuit. (FIG. 17B) The lactose and arabinose internal inducer concentrations. (FIG. 17C) The reporter protein levels. (FIG.17D) The emergent robotic behavior within the arena.

FIGS. 18A-18D show stochasticity in a balanced toggle with 0% Transcription and 5% Translation variance. (FIG. 18A) The EFM signal for the stochastic circuit. (FIG. 18B)

 

The lactose and arabinose internal inducer concentrations. (FIG. 18C) The reporter protein levels. (FIG.18D) The emergent robotic behavior within the arena.

FIGS. 19A-19D show Stochasticity in a Balanced Toggle with 5% Transcription and 0% Translation variance. (FIG. 19A) The EFM signal for the stochastic circuit. (FIG. 19B) The lactose and arabinose internal inducer concentrations. (FIG. 19C) The reporter protein levels. (FIG.19D) The emergent robotic behavior within the arena.

FIGS.20A-20D show stochasticity in a balanced toggle with 5% transcription and 5% translation variance. (FIG. 20A) The EFM signal for the stochastic circuit. (FIG. 20B) The lactose and arabinose internal inducer concentrations. (FIG. 20C) The reporter protein levels. (FIG.20D) The emergent robotic behavior within the arena.

FIGS. 21A-21E show a bioinspired decision architecture for a Robot. (FIG. 21A) A physical robot was built to act as a host processing hub. (FIG. 21B) By employing the techniques of synthetic biology, logic-based gene regulatory networks may be transformed into E. coli, causing cells to produce fluorescent reporter proteins when cells are exposed to different input chemical inducers, such as arabinose and lactose. (FIG.21C) The fluorescent signal proteins produced by the simulated E. coli population can be optically detected by an epifluorescent microscope and processed into an analogue signal. Therefore, we can treat them as a detectible signal output. (FIG.21D) This signal can then be converted into a digital epiflourescent microscope value (EFM value) by setting analogue thresholds; the EFM value serves as the input for a finite state machine that toggles robot subroutines, resulting in actuation of a physical robot. (FIG. 21E) By incorporating onboard, integrated computer vision, this design allows us to engineer methods for environmental problem solving using a microbiome processing hub and a host processing hub in parallel.

FIGS. 22A-22D show a master/slave architecture resulting in biased random walk robot behavior driven by a mutually repressible gene network. (FIG. 22A) A robot was engineered to move with distinct periods of “runs” (forward locomotion) and“tumbles” (rotational locomotion). (FIG. 22B) This behavior was the result of a master/slave control structure linking signal proteins from the microbiome processing hub to distinct robot sub- routines. Under these conditions, the production of signal protein one (SP1) above a threshold activates a robot-based biased random walk towards a green target. (FIG. 22C) The production of signal proteins was governed by a mutually repressible, regulatory gene network, and simulated within the in silico microbiome processing hub. This regulatory network is bistable and balanced with expression levels of the repression proteins functionally equal. (FIG. 22D) This gene network allows a temporary pulse of an input chemical (lactose) to cause a sustained, elevated production of a signal protein. A plot of SP1 is shown. The red circles correspond to times when lactose is inputted into the microbiome processing hub.  

FIGS 23A-23D show run and tumble architecture resulting in biased random walk robot behavior driven by a weighted mutually repressible gene Network. (FIG. 23A) A robot was engineered to move with distinct periods of“runs” (forward locomotion) represented by cyan dashes and“tumbles” (rotational locomotion) represented by purple arrows. (FIG.23B) This behavior was the result of a hybrid control structure linking cell-synthesized signal proteins, SP1 and signal protein two (SP2), to distinct robot sub-routines. Under these conditions, the production of SP1 above a threshold caused the robot to move forward, or “run.” When SP2 was produced above a threshold, the robot was told to rotate, or“tumble”, and deliver a small pulse of lactose to the simulated cells within the microbiome processing hub. Additionally, a pulse of lactose was delivered to the microbiome processing hub if the size of the target grew between a time step, as processed by the onboard computer vision system. (FIG. 23C) The production of signal proteins was governed by a biased mutually repressible regulatory gene network modeled and simulated within the in silico microbiome processing hub. This gene network is similar to the one in FIGS. 21A-21E, except for an upregulation of the LacI repression protein on one side of the gene circuit. (FIG. 23D) This regulatory network is biased. A transient pulse of lactose causes a temporary elevation of signal protein one. However, without more lactose added to the system, the bias of the network lowers the production of SP1 until the toggle flipped and SP2 is produced. A plot of the signal proteins is shown. The red circles correspond to times when lactose is pulsed into the microbiome processing hub.

FIGS. 24A-24D show Integrated Perception Architecture Causes Biased Random Walk Robot Behavior Driven by Two Self-Repressing Gene Network. (FIG.24A) A robot was engineered to move with distinct periods of“runs” (forward locomotion) represented by cyan dashes and“tumbles” (rotational locomotion) represented by purple arrows. (FIG. 24B) This behavior was the result of a hybrid control structure linking cell synthesized signal proteins to distinct robot sub-routines. Under these conditions, if SP1 was produced at higher levels than SP2, the robot will walk forward, or“run.” When SP2 was produced at higher levels than SP1, the robot was told to rotate, or“tumble”. Additionally, a chemical input pulse of AHL inducer was delivered to the microbiome processing hub in an amount proportional to the size of the target, as determined by the onboard camera and computer vision. (FIG.24C) The production of signal proteins was governed by a naturally oscillating gene network modeled and simulated within the in silico microbiome processing hub. (FIG. 24D) This regulatory network caused an oscillatory pattern. However, when AHL is added to the system, there is an elevated production of SP1. This elevation continues as long as there is an increasing amount of AHL added, otherwise negative feedback will attenuate the SP1 production. A plot of the signal proteins is shown. The orange rectangles correspond to the time and size of the AHL pulse delivered to the microbiome processing hub.  

FIGS. 25A-25D demonstrate that mutually repressible gene circuit drives alternating resource selection. By using the same gene circuit and control architecture depicted in FIGS. 22A-22D, the robot can be simulated to alternate between different targets. (FIG. 25A) The robot is first observed executing a biased random walk towards the green target. Upon arriving at the green target, the chemical input pulse of arabinose inducer is delivered, flipping the mutually repressible gene circuit and the robot executes a biased random walk towards the red target. (FIG. 25B) A plot of the signal proteins is show. The red circle corresponds to times when arabinose is pulsed into the microbiome processing hub. (FIG. 25C) By modifying the robot’s locomotion subroutine, we can demonstrate how the same gene network may be used to engineer a direct target seeking behavior. (FIG.25D) A plot of the corresponding signal proteins is shown. The green/red circles correspond to times when arabinose/lactose is pulsed into the microbiome processing hub.

FIGS. 26A-26D demonstrate that biased mutually repressible gene circuit drives preferential target selection. (FIG. 26A) A robot is observed moving in a pattern similar to preferential resource selection. The robot spends the majority of its time seeking the green target, but when close to red target it temporarily searches for the red target. (FIG.26B) The underlying control architecture is similar to that found in FIGS. 22A-22D. However, an additional condition was added so that the robot delivers a pulse of arabinose to the microbiome processing hub when the robot is in proximity to a red target. (FIG. 26C) A biased, mutually repressible gene network, identical to that used in FIGS. 23A-23D, was simulated. (FIG. 26D) A plot of the corresponding signal proteins is shown. The green/red circles correspond to times when arabinose/lactose is pulsed into the microbiome processing hub. Note that the arabinose is given when the robot is within the black dashed line and lactose is given to the robot makes contact with a red target.

FIG. 27 shows a figure demonstrating that mutable biological components can optimize robotic behavior. The behavior of the robot can be modified by altering the strength of the bias within the mutually repressible genetic circuit.250 live robot trials were conducted (25 trials for each RBS strength value) using the same experimental design and genetic topology presented in FIGS. 26A-26D. The results show how variations in the ribosome binding site strength, an easily mutable component, can shift robot behavior.

DETAILED DESCRIPTION Before the present disclosure is described in greater detail, it is to be understood that this disclosure is not limited to particular embodiments described, and as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting.  

Where a range of values is provided, it is understood that each intervening value, to the tenth of the unit of the lower limit unless the context clearly dictates otherwise, between the upper and lower limit of that range and any other stated or intervening value in that stated range, is encompassed within the disclosure. The upper and lower limits of these smaller ranges may independently be included in the smaller ranges and are also encompassed within the disclosure, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the disclosure.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. Although any methods and materials similar or equivalent to those described herein can also be used in the practice or testing of the present disclosure, the preferred methods and materials are now described.

All publications and patents cited in this specification are herein incorporated by reference as if each individual publication or patent were specifically and individually indicated to be incorporated by reference and are incorporated herein by reference to disclose and describe the methods and/or materials in connection with which the publications are cited. The citation of any publication is for its disclosure prior to the filing date and should not be construed as an admission that the present disclosure is not entitled to antedate such publication by virtue of prior disclosure. Further, the dates of publication provided could be different from the actual publication dates that may need to be independently confirmed.

As will be apparent to those of skill in the art upon reading this disclosure, each of the individual embodiments described and illustrated herein has discrete components and features which may be readily separated from or combined with the features of any of the other several embodiments without departing from the scope or spirit of the present disclosure. Any recited method can be carried out in the order of events recited or in any other order that is logically possible.

Embodiments of the present disclosure will employ, unless otherwise indicated, techniques of molecular biology, microbiology, nanotechnology, organic chemistry, biochemistry, botany, synthetic biology, chemical engineering, bioengineering, electrical engineering, mechanical engineering, software engineering and the like, which are within the skill of the art. Such techniques are explained fully in the literature.

Definitions

As used herein, "about," "approximately," and the like, when used in connection with a numerical variable, generally refers to the value of the variable and to all values of the

 

variable that are within the experimental error (e.g., within the 95% confidence interval for the mean) or within +/- 10% of the indicated value, whichever is greater.

The term“biocompatible”, as used herein, refers to a material that along with any metabolites or degradation products thereof that are generally non-toxic to the recipient and do not cause any significant adverse effects to a microorganism.

As used herein,“deoxyribonucleic acid (DNA)” and“ribonucleic acid (RNA)” generally refer to any polyribonucleotide or polydeoxribonucleotide, which may be unmodified RNA or DNA or modified RNA or DNA. RNA may be in the form of a tRNA (transfer RNA), snRNA (small nuclear RNA), rRNA (ribosomal RNA), mRNA (messenger RNA), anti-sense RNA, RNAi (RNA interference construct), siRNA (short interfering RNA), or ribozymes.

As used herein,“DNA molecule” includes nucleic acids/polynucleotides that are made of DNA.

As used herein,“expression” refers to the process by which polynucleotides are transcribed into RNA transcripts. In the context of mRNA and other translated RNA species, “expression” also refers to the process or processes by which the transcribed RNA is subsequently translated into peptides, polypeptides, or proteins.

As used herein, “heterogeneous” can refer to a population of molecules or microorganisms in which at least two of the molecules or microorganisms are different from each other.

As used herein, “homogenous” can refer to a population of molecules or microorganisms in which all the molecules or microorganisms are identical with one another.

As used herein,“microbiome” can refer to a population of microorganisms that inhabit a specific environment, and thus can create a mini-ecosystem. It can also be used herein to refer to a population of bacteria that inhabit a specific environment. A“microbiome” as used herein can contain a heterogeneous or homogenous population of microorganisms and/or bacteria.

As used herein,“optically active protein” can refer to a protein that can emit wavelengths of light. Such proteins include but are not limited to fluorescent proteins (e.g. green fluorescent proteins, red fluorescent proteins, yellow fluorescent proteins, blue fluorescent proteins and any variant thereof).

As used herein,“plasmid” as used herein refers to a non-chromosomal double- stranded DNA sequence including an intact“replicon” such that the plasmid is replicated in a host cell.

As used herein,“protein” as used herein refers to a large molecule composed of one or more chains of amino acids in a specific order. The term protein is used interchangeable with“polypeptide.” The order is determined by the base sequence of nucleotides in the gene

 

coding for the protein. Proteins are required for the structure, function, and regulation of the body’s cells, tissues, and organs. Each protein has a unique function.

As used herein,“robotic device” can refer to any mechanical artificial agent, typically an electromechanical machine that can be guided or otherwise controlled to perform a function by an application or other processing or electrical circuitry.“Robotic device” can perform functions automatically in response to a command received from an application or other processing or electrical circuitry.

As used herein, "substantially all" can mean that more than about 80 percent of all microorganisms present in the microbiome, more preferably more than about 85%, 90%, 95%, and 99%.

As used herein,“substantially homogenous” can mean that more than about 80 percent, and preferably, more than about 85%, 95%, and 99% of all microorganisms in the microbiome are the same family, genus, and/or species of microorganisms. As used herein “substantially homogenous” can also refer, depending on the context used, that more than about 80 percent, and preferably, more than about 85%, 95%, and 99% of all microorganisms contain the same synthetic gene circuit(s).

As used herein,“substantially pure cell population” can refer to a population of cells having a specified cell marker characteristic and differentiation potential that is about 50%, preferably about 75-80%, more preferably about 85-90%, and most preferably about 95% of the cells making up the total cell population. Thus, a "substantially pure cell population" can refer to a population of cells that contain fewer than about 50%, preferably fewer than about 20-25%, more preferably fewer than about 10-15%, and most preferably fewer than about 5% of cells that do not display a specified marker characteristic and differentiation potential under designated assay conditions.

As used herein,“tenability,”“tunable” and the like can refer to the ability to adjust the activity level of a synthetic gene circuit or component thereof, including but not limited to the strength of gene expression from a promoter.

As used herein,“toggle switch” can refer to a circuit that can exist in one of two stable states and can be switched (toggeled) between the two states by a defined input. Discussion

An organism’s evolutionary fitness is determined by how well it utilizes environmental metabolites. For constituents of the microbiome - the microorganisms associated with the animal body – their environment is a product of their host’s physiology. Yet, these commensal microbes also play a critical role in governing the health and behavior of their hosts. The effects include impacting host metabolism, perturbing host hormone regulation and changing the host’s affinity for disease (Backhed, ed al., Proc. Natl. Acad. Sci. USA 101,  

15718-15723, 2004; Markle, et al., Science 339, 1084-1088, 2013; Tlaskalova-Hogenova, et al., Immunol. Lett. 93, 97-108, 2004). These interactions can even regulate complex animal behavior. For example, recent studies found that commensal Lactobacillus plantarum can affect the mating behavior of their Drosophila melanogaster hosts, and that microbiome density can directly influence anxiety, and by extension, motility in mice (Sharon, et al., Proc. Natl. Acad. Sci. USA 107, 20051-20056, 2010; Neufeld, et al., Neurogastroenterol. Motil.23, 255-264, 2011).

While examining the physiology of commensal microbes is important, macro-scale host behaviors must also be better understood to elucidate host-microbiome interactions. Biomimetic approaches give a robust toolset for analyzing animal behavior. These robots can provide quantifiable, minimal systems representative of natural phenomenal and are useful for scientific inquiry. Further, such devices can be useful for environmental screening. As such there is a need for improved biomimetic devices for understanding complex microbiome-host interactions and for detection of environmental compounds.

With that said, described herein are biomimetic systems that can contain a microbiome that can have at least one genetically modified bacterium having a synthetic, gene circuit, where the microbiome can be coupled to a sensor that can be coupled to a computing device that can have a processor capable of executing an application that can cause the processing circuitry to at least receive a signal from the sensor and send a command to a robotic device to execute a function. Also provided herein are methods of controlling a robotic device by growing a microbiome that can have at least one genetically modified bacterium having a synthetic gene circuit, sensing a biological activity of the microbiome, generating a signal in response to the sensed biological activity, and transmitting a command to a robotic device to execute a function based at least in part upon the output voltage.

Other systems, methods, features, and advantages of the present disclosure will be or become apparent to one having ordinary skill in the art upon examination of the following drawings, detailed description, and examples. It is intended that all such additional compositions, compounds, methods, features, and advantages be included within this description, and be within the scope of the present disclosure.

Biomimetic Systems

Provided herein are biomimetic systems 1000 that can contain a microbiome 1010, a sensor 1020, processing circuitry 1030 that can have a processor and an application that can be executed by the processor (e.g. FIGS. 8A-8C). The microbiome can contain a genetically engineered bacterium. The genetically engineered bacterium can contain at least one synthetic gene circuit that can be responsive to an inducer. The microbiome 1010 can be contained in a chemostat or other suitable container 1050. The sensor 1020 can be  

coupled to the microbiome 1010. The sensor 1020 can be configured to produce an output signal that can correspond to one or more biological activity of one or more bacterium or product thereof of the microbiome 1010 and/or the microbiome 1010 as a substantial whole or a product thereof that is detected by the sensor. The processing circuitry 1030 can contain a processor and a memory. The processing circuitry 1030 can be coupled to the sensor 1020. The application can contain machine readable instructions that can be stored in the memory and that, when executed by the processor, can cause the processing circuitry to at least receive the output signal from the sensor and provide a command to a robotic device 1040 to execute a function based at least in part upon the output signal. The various components of the biomimetic systems 1000 provided herein can exists in various configurations as will be appreciated by those of ordinary skill in the art in view of the descriptions of the biomimetic systems and components thereof provided herein. Some non- limiting examples of different configurations of the biomimetic systems 1000 are provided in FIGS. 8A-8C and discussed elsewhere herein. With a general understanding of the biomimetic systems provided herein in mind, the biomimetic systems and components thereof will be now be discussed in greater detail.

Microbiome

The biomimetic systems provided herein can contain a microbiome. The microbiome can contain one or more microorganisms. The microbiome can be homogenous or heterogeneous as to the type (e.g. genus and/or species) of microorganism present in the microbiome. In some embodiments, the microbiome contains one or more types of bacteria. In some embodiments, the microbiome contains bacterium and at least one other microorganism from another classification family (e.g. viruses and yeast). The microbiome can be synthetic. In other words, the microbiome can be generated by combining the microorganisms in known quantities to result in a synthetic microbiome. In other embodiments, the microbiome can be obtained by obtaining a sample from a natural environment (such as a stream, pond, or from an organism). The microbiome can be cultivated using techniques known to one of ordinary skill in the art.

The microbiome, whether cultivated by combination from established and/or genetically modified strains and variants of microorganisms, by sampling from a native environment, or some combination thereof, at least one of the microorganisms, such as, but not limited to a bacterium, can be genetically engineered to contain a synthetic gene circuit, which can be responsive to an inducer. Methods of genetically modifying bacterium and other microorganisms contemplated here are well-established in the art and generating a genetically modified bacterium or other microorganism can be accomplished using techniques known to one of ordinary skill in the art.

 

In some embodiments the microbiome can include or only contain one or more bacterial species or strains thereof. Suitable bacteria that can genetically modified and/or be part of the microbiome include, but are not limited to, bacteria of the genus Escherichia (e.g. E. coli, including but not limited to all strains derived from the K-12 parent strain (e.g. MG1655) bacteria of the genus Vibrio (e.g., Vibrio cholerae), bacteria of the genus Aliivibrio (e.g., Aliivibrio fischeri), bacteria of the genus Bacillus (e.g. Bacillus subtilis), bacteria of the genus Lactobacillus (e.g. Lactobacillus acidophilus), bacteria of the genus Pseudomonas (e.g. Pseudomonas aeruginosa), bacteria of the genus Ralstonia (e.g. Ralstonia eutropha), and bacteria of the genus Staphylococcus (e.g. Staphylococcus aureus). In some embodiments the microbiome can include or only contain one or more yeast species or strains thereof. Suitable yeast include but are not limited to yeast of the genus Saccharomyces (e.g. S. cerevisiae).

At least one of the microorganisms in the microbiome can have at least one synthetic gene circuit. In some embodiments, substantially all the microorganisms in the microbiome contain at least one synthetic gene circuit. The synthetic gene circuit(s) can be configured as a switch, a bistable switch, a toggle switch, an oscillator, a repressilator, counter, anticipator, learner, kill switch, quorum sensor (sender or receiver), two-way signaling system, and- gates, nor-gate, nand-gate, inverter, or-gate, engineered ecosystem circuits, single invertase memory modules, analog-to-digital converter, digital-to-analog converter or any permissible combination thereof. The synthetic gene circuit can be tunable and thus allow for modulation of expression of one or more reporter genes. In some embodiments, the synthetic gene circuit can be designed using a program such as GenoCAD, Clotho framework, or j5. The synthetic gene circuit can contain one or more reporter genes. Suitable reporter genes are generally known in the art. Such reporter genes include, but are not limited to, optically active proteins (e.g. green fluorescent protein and variants thereof (e.g. eGFP), red fluorescent protein and variants thereof (e.g. mCherry)), lacZ (produces beta-galatosidase), cat (produces chloramphenicol acetyltransferase), beta-lactamase, and other antibiotic resistance genes.

The reporter genes can be operatively coupled to one or more transcriptional control elements. As used herein,“transcriptional control element” can refer to any element of the synthetic gene circuit, including proteins, DNA, RNA, or other molecules that can, either alone or in conjunction with other elements of the synthetic gene circuit, stimulate and/or repress the transcription of one or more reporter genes within the synthetic gene circuit. Such transcriptional control elements will be apparent to those in the art and include, but are not limited to operons and components thereof, bacterial repressors, eukaryotic promoters and elements therein, DNA binding proteins, signaling molecules, riboregulators, toe-hold switches, siRNA, CRISPR/Cas9, TALENs, Zinc-Finger Nucleases (ZFNs), and the like.  

Others, in view of those presented here, will be apparent to those of ordinary skill in the art. In some embodiments, the synthetic gene circuit can include one or more components (or all of the components) of the Lac operon and/or the Tet repressor.

The transcriptional control element(s) can be responsive, either directly (e.g. by direct binding) or indirectly (e.g. through a signaling cascade) to an inducer. An inducer can be electromagnetic waves (including, but not limited to, light), a temperature change (e.g. heat- shock), a chemical, a small molecule, an organism or microorganism or component thereof. In some embodiments and inducer can be IPTG, aTC, lactose, arabinose, acyl-homoserine lactones (e.g. AHL, 3O-C12).

The synthetic gene circuits can be designed and generate using techniques provided herein and other techniques generally known to one of ordinary skill in the art. They can be generated as plasmids (bacterial or viral), naked DNA constructs, artificial chromosomes, and in any other suitable format. The synthetic gene circuits can be integrated into a host microorganism, such as a bacterium, using techniques generally known to the skilled artisan. In other words, the host microorganism can be transformed with the synthetic gene circuit. The synthetic gene circuit(s) can be stably or transiently expressed within the microorganism, as the terms would be understood in the art.

The microbiome can be contained in a chemostat. The chemostat can be a micro- chemostat. In some embodiments the chemostat can include a microfluidic channel that can contain the microorganisms of the microbiome. The media or other fluid in the chemostat can be controlled such that the microorganisms are maintained in a specific growth phase. In some embodiments, the microorganisms can be maintained in an exponential phase, stationary phase, or lag phase of growth by controlling the fluid composition and fluid in the chemostat. In some embodiments, over a period of time, the microorganisims can be maintained such that they change from one phase of growth to another at a desired time by controlling the fluid composition and fluid in the chemostat.

Sensors

The biomimetic systems described herein can include one or more sensors. The sensors can be coupled to the microbiome and can be configured to produce an output signal that can correspond to a biological activity of the microbiome detected by the sensor. It will be appreciated that the term“biological activity” as used herein can refer to any biologic activity of the microbiome, whether the result of endogenous gene activity or synthetic gene circuit activity. This includes, but is not limited to, protein and other molecule expression and secretion, reporter gene expression and/or activity, growth rate, metabolism changes, and other similar cellular activities. In some embodiments, the sensor can be directly in contact with the microorganisms of the microbiome and/or coupled to the chemostat such that the sensor is at last contacts the fluid within the chemostat. In other  

embodiments the sensor can be coupled to an intermediate device, such as a microscope, that facilitates collection of characteristic information (e.g. light emitted from the microbiome) and transport of that characteristic information to the sensor. It will be appreciated that in some cases the sensor and the intermediate device are inseparable (i.e. integrated within the intermediate device). The sensor can be fluidicaly, optically, wirelessly, electrically, chemically, and/or biologically coupled to the microbiome, chemostat, and/or component thereof.

The sensor(s) can be configured to directly measure a biologic activity of a microorganism, microorganism population within the microbiome, the microbiome, and/or the microbiome environment (e.g. the fluid in the chemostat) or can be configured to indirectly detect a biological activity of a microorganism, microorganism population within the microbiome, the microbiome, and/or the microbiome environment (e.g. the fluid in the chemostat). The sensor(s) can be configured to convert an energy (e.g. light) input into an output signal, such as a voltage, impedance, sound, light, or other signal. The sensor(s) can be biocompatible.

In some embodiments, the sensor(s) can detect light or other signal (such as a protein product or activity of a protein product) produced from one or more of the reporter gene(s) that can be contained in the synthetic circuit. In some embodiments where the reporter gene is an optically active protein, the sensor(s) can detect light (such as fluorescent light) emitted from the optically active protein.

Suitable sensors will be apparent to one of ordinary skill in the art and can include, but are not limited to, accelerometers, hygrometer, microphones, chemical sensors, biomolecule sensors, electrochemical gas sensors, electrolyte-insulator-semiconductor sensors, fluorescent chloride sensors, fluorescence resonance energy transfer (FRET)- based sensors, holographic sensors, hydrocarbon dew point sensors, surface acoustic wave sensors, nondispersive infrared sensors, ion selective electrodes, olfactometers, optical sensors, photodiodes, pellistors, potentiometric sensors, zinc oxide nanorod sensors, current sensors, Daly detectors, galvanometers, Hall effect sensors, Hall probes, magnetometers, magnetic field sensors, voltmeters, flow sensors, mass flow sensors, cloud chambers, gyroscopes, altimeters, auxanometers, capacitive displacement sensors, gravimeters, inclinometers, integrated circuit piezoelectric sensors, laser surface velocimeters, linear variable differential transformers, odometers, photoelectric sensors, piezoelectric sensors, position sensors, rate sensors, rotary encoders, rotary variable differential transformers, stretch sensors, ultrasonic thickness gauges, variable reluctance sensors, velocity receivers, colorimeters, contact image sensors, electro-optical sensors, infrared sensors, kinetic inductance detectors, light emitting diode sensors, light addressable potentiometric sensors, fiber optic sensors, optical position sensors, photo detectors, phototransistors,  

photoionization detectors, photo-electric switches, scintillometers, single-photon avalanche diodes, superconducting nanowire single-photon detectors, transition edge sensors, visible light photon counters, wavefront sensors, barographs, barometers, densitometers, pressure sensors, tactile sensors, bhangmeters, hydrometers, force sensors, level sensors, torque sensors, viscometers, strain gauges, bolometers, microbolometers, bimetallic strips, gardon gauges, heat flux sensors, thermometers, thermistors, pyrometers, proximity sensors, reed switches, BioMEMs and BioMEM based sensors, and/or photoelastic sensors.

Processing Circuitry and Applications Executed by the Processing Circuitry The biomimetic system provided herein can contain processing circuitry. The processing circuitry can contain a processor and a memory. It will be understood that where reference is made in this application to processing circuitry, it is implied that the processing circuitry contains a processor and memory. The processing circuitry can be coupled to the sensor. The processing circuitry can be electrically, optically, or wirelessly coupled to the sensor. The biomimetic system provided herein can contain an application that can contain machine readable instructions, such as a software program, stored in the memory, that, when executed by the processor, can cause the processing circuitry to at least receive the output signal from the sensor and provide a command to a robotic device to execute a function based at least in part upon the output signal. The application can further determine a

Stored in the memory can be both data and several components that are executable by the processor, which in some embodiments, can make up the whole or part of the application. In particular, stored in the memory and executable by the processor can be applications capable of implementing the communication of sensor data and operation of a robotic device as discussed herein, and potentially other applications. Also stored in the memory can be a data store including, e.g., collected sensor data and other data that can be received from the robotic device such as position, speed, etc. In addition, an operating system may be stored in the memory and executable by the processor. It is understood that there may be other applications that are stored in the memory and are executable by the processor as can be appreciated.

Where any component discussed herein is implemented in the form of software, any one of a number of programming languages may be employed such as, for example, C, C++, C#, Objective C, Java^®, JavaScript^®, Perl, PHP, Visual Basic^®, Python^®, Ruby, Delphi^®, Flash^®, or other programming languages. A number of software components are stored in the memory and are executable by the processor. In this respect, the term "executable" can refer to a program file that is in a form that can ultimately be run by the processor. Examples of executable applications or programs may be, for example, a compiled program that can be translated into machine code in a format that can be loaded into a random access portion  

of the memory and run by the processor, source code that may be expressed in proper format such as object code that is capable of being loaded into a random access portion of the memory and executed by the processor, or source code that may be interpreted by another executable program to generate instructions in a random access portion of the memory to be executed by the processor, etc. An executable program may be stored in any portion or component of the memory including, for example, random access memory (RAM), read-only memory (ROM), hard drive, solid-state drive, USB flash drive, memory card, optical disc such as compact disc (CD) or digital versatile disc (DVD), floppy disk, magnetic tape, or other memory components.

The memory can be defined herein as including both volatile and nonvolatile memory and data storage components. Volatile components are those that do not retain data values upon loss of power. Nonvolatile components are those that retain data upon a loss of power. Thus, the memory can contain, for example, random access memory (RAM), read- only memory (ROM), hard disk drives, solid-state drives, USB flash drives, memory cards accessed via a memory card reader, floppy disks accessed via an associated floppy disk drive, optical discs accessed via an optical disc drive, magnetic tapes accessed via an appropriate tape drive, and/or other memory components, or a combination of any two or more of these memory components. In addition, the RAM may comprise, for example, static random access memory (SRAM), dynamic random access memory (DRAM), or magnetic random access memory (MRAM) and other such devices. The ROM may include, for example, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or other like memory device.

Also, the processor can represent multiple processors and the memory can represent multiple memories that operate in parallel processing circuits, respectively. In such a case, the local interface can be an appropriate network that facilitates communication between any two of the multiple processors, between any processor and any of the memories, or between any two of the memories, etc. The processor can be of electrical or of some other available construction (e.g. optical).

Although portions of the application(s) and other various systems described herein can be embodied in software or code executed by general purpose hardware, as an alternative the same can also be embodied in dedicated hardware or a combination of software/general purpose hardware and dedicated hardware. If embodied in dedicated hardware, each can be implemented as a circuit or state machine that employs any one of or a combination of a number of technologies. These technologies can include, but are not limited to, discrete logic circuits having logic gates for implementing various logic functions upon an application of one or more data signals, application specific integrated circuits  

having appropriate logic gates, or other components, etc. Such technologies are generally well known by those skilled in the art and, consequently, are not described in detail herein.

The application(s) can contain program or machine readable instructions to implement logical function(s) and/or operations of the system. The program instructions or machine readable instructions can be embodied in the form of source code that comprises human-readable statements written in a programming language or machine code that comprises numerical instructions recognizable by a suitable execution system such as a processor in a computer system or other system. The machine code may be converted from the source code, etc. If embodied in hardware, each block may represent a circuit or a number of interconnected circuits to implement the specified logical function(s).

Also, any logic or application described herein that contains software or code can be embodied in any non-transitory computer-readable medium for use by or in connection with an instruction execution system such as, for example, a processor in a computer system or other system. In this sense, the logic may include, for example, statements including instructions and declarations that can be fetched from the computer-readable medium and executed by the instruction execution system. In the context of the present disclosure, a "computer-readable medium" or“machine-readable medium” can be any medium that can contain, store, or maintain the logic or application described herein for use by or in connection with the instruction execution system.

The computer-readable medium or machine readable medium can comprise any one of many physical media such as, for example, magnetic, optical, or semiconductor media. More specific examples of a suitable computer-readable medium or machine-readable medium include, but are not limited to, magnetic tapes, magnetic floppy diskettes, magnetic hard drives, memory cards, solid-state drives, USB flash drives, or optical discs. Also, the computer-readable medium may be a random access memory (RAM) including, for example, static random access memory (SRAM) and dynamic random access memory (DRAM), or magnetic random access memory (MRAM). In addition, the computer-readable medium may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or other type of memory device.

It will be appreciated that when devices or components are described in this application as being optically or wirelessly coupled that the transmitting device and the receiving device contain the appropriate transmitters and/or receivers to transmit and receive a signal. Some or all of the transmitter, receiver, antenna (if for wireless connectivity), processors, controls, analog-to-digital convertors, digital-to-analog convertors, and memory blocks or any other components can be on the cellular interfacing array chip, substrate, printed circuit board, cell culture container, mesh, external devices, or combination thereof.  

Robotic Devices

The biomimetic systems provided herein can include a robotic device. The robotic device can be directly or indirectly coupled to the microbiome, chemostat, sensor processing circuitry, and/or application(s) described herein. In other words, one or more of the other components of the biomimetic system described herein can be coupled to in a suitable manner (electrically, physically, fluidicaly, operatively, wirelessly, optically or otherwise) coupled to the robotic device. In some embodiments, all the components described herein are coupled to the robotic device such that all the components are housed on and/or contained within the robotic device (see e.g. FIG. 8). In other embodiments at least one of the components is separate from the robotic device (see e.g. FIG. 8B). For example, in some embodiments, the microbiome can be contained in a chemostat and coupled to a sensor that can wirelessly deliver information processing circuitry and applications that are housed on or contained within the robotic device.

In other embodiments, all components are separate and remote from the robotic device (see e.g. FIG.8C). The processing circuitry and/or application can be coupled to and provide commands to the robotic device, for example, wirelessly. In some of these embodiments the microbiome, sensor, processing circuitry, and application can be contained in the same location. In other embodiments, the microbiome and sensor can be kept separate from the processing circuitry. Other implementations will be appreciated in view of the description provided here and are within the spirit and scope of this disclosure.

Operation and Uses of the Biomimetic System

In operation, a microbiome described herein can be grown and maintained in a suitable environment. In some embodiments, the microbiome is grown such that substantially all of the microorganisms in the microbiome are at and/or maintained at a desired growth phase. In some embodiments, the desired growth phase can be the exponential growth phase. The microbiome can be grown in a chemostat. The microbiome can be exposed to an inducer. The inducer can stimulate or repress a biologic activity (including but not limited to activation and/or repression of a synthetic gene circuit or component thereof).

A sensor can then detect a desired biological activity of the microbiome. In embodiments where the microbiome contains a synthetic gene circuit containing a reporter gene, the sensor can detect the presence and/or activity of the reporter gene. In some cases where multiple biologic activities are to be detected, either individually or simultaneously, one or more sensors can be used to detect the biologic activities. In some embodiments where a light energy from a reporter protein produced the microbiome or microbiome environment is desired to be sensed, a microscope can be used to collect and image the light. The microscope can contain or be coupled to the sensor that can detect the light.  

The sensor can convert the detected input into an output signal that can be transmitted to the processing circuitry. In response to the received output signal an application can be executed that can, in response to an input, determine and send a command to the robotic device to perform a function. In some embodiments, the application can determine a characteristic of the output signal and then, in response to the determined characteristic of the output signal, determine and send a command to the robotic device to perform a function. The function can be any operative function that can be performed by a particular robotic device described herein, including but not limited to moving, stopping, starting, lifting objects, suctioning (e.g. sampling) environmental fluids, ejecting fluids into the environment, dropping objects, turning on and off a light, signaling to other robots, modifying robot body (e.g. changing physical form), taking images, analyzing images, taking video, processing video, communicating with central server, placing explosives, removing explosives, initiating onboard reactions, incubating and monitoring living cells, incubating and monitoring cell-free reaction systems, self-destructing, and/or any permissible combination thereof.

In some embodiments, the microbiome can evolve in response to particular inducers and thus can result in controlling the robot in developed manner that is not necessarily preprogrammed in the application or other portion of the processing circuitry.

In these ways, a microbiome provided herein can control the function of a robotic device and/or evaluate the host-microbiome interaction. The biomimetic systems provided herein can thus be used to evaluate microbiome-host interactions. The biomimetic systems provided herein can be used to evaluate the effect of a compound present in the environment on the microbiome and/or microbiome-host interaction. In some embodiments, the robotic device can be pre-programmed to troll an environment and sample unknown inducers. The microbiome can respond (or not) to the sampled unknown inducer and control the response of the robotic device accordingly, thus providing feedback on the nature of the sampled unknown inducers, such as its toxicity on the microbiome or host. Other potential uses for the biomimetic systems provided herein not expressly stated can be within the understanding of one of ordinary skill in the art and will be appreciated to be within the sprit and scope of this disclosure. EXAMPLES Now having described the embodiments of the present disclosure, in general, the following Examples describe some additional embodiments of the present disclosure. While embodiments of the present disclosure are described in connection with the following examples and the corresponding text and figures, there is no intent to limit embodiments of  

the present disclosure to this description. On the contrary, the intent is to cover all alternatives, modifications, and equivalents included within the spirit and scope of embodiments of the present disclosure. Example 1.

Introduction:

An organism’s evolutionary fitness is determined by how well it utilizes environmental metabolites. For constituents of the microbiome - the microorganisms associated with the animal body – their environment is a product of their host’s physiology. Yet, these commensal microbes also play a critical role in governing the health and behavior of their hosts. The effects include impacting host metabolism, perturbing host hormone regulation and changing the host’s affinity for disease (Backhed, ed al., Proc. Natl. Acad. Sci. USA 101, 15718-15723, 2004; Markle, et al., Science 339, 1084-1088, 2013; Tlaskalova-Hogenova, et al., Immunol. Lett. 93, 97-108, 2004). These interactions can even regulate complex animal behavior. For example, recent studies found that commensal Lactobacillus plantarum can affect the mating behavior of their Drosophila melanogaster hosts, and that microbiome density can directly influence anxiety, and by extension, motility in mice (Sharon, et al., Proc. Natl. Acad. Sci. USA 107, 20051-20056, 2010; Neufeld, et al., Neurogastroenterol. Motil.23, 255-264, 2011).

These correlations likely result from inter-kingdom communication through biochemical signaling (Hughes, et al., Nat Rev Micro 6, 111-120 (2008). Furthermore, these communication motifs are present as relationships between consortia of commensal microbes and their host, not merely as interactions between a single microbial species and its host. Microbial consortia - consisting of multiple species and intertwined biochemical networks, allow for network complexity, as do spatial variations in the host ecosystem, rendering host-microbiome interactions difficult to fully understand and model (Walter, et al., Annu. Rev. Microbiol.65, 411-428, 2011; Greenblum, et al., Curr. Opin. Biotechnol.24, 810- 820, 2013).

A variety of scientific tools exist to help explore complicated biological systems. Synthetic biology has generated multiple tools that have been used to probe and program cellular behaviors over the past fifteen years. The field was launched in 2000 by reports of the first synthetic biological networks– or engineered gene circuits– that functioned as memory or oscillatory modules in cells (Brophy, et al., Nat. Methods 11, 508-520, 2014; Gardner, et al., Nature 403, 339-342, 2000; Elowitz, et al., Nature 403, 335-338, 2000). Inspired by electrical engineering, these circuits have expanded to include other modules such as logic gates, timers, counters, and simple analog computers (Anderson, et al., Mol Syst Biol 3, 133, 2007; Ellis, et al., Nat. Biotechnol. 27, 465-471, 2009; Friedland, et al.,  

Science 324, 1199-1202, 2009; Daniel, et al., Nature 497, 619-623, 2013). These behaviors are programmed into DNA and then uploaded into cells. The resulting synthetic networks can then interface with endogenous networks within the same cell, organism, or commensal host to reprogram behavior (FIGS.1A-1C).

Progress in synthetic biology includes the creation of genetic component libraries as well as computational tools, giving researchers the ability to simulate analytical models of cellular response in silico prior to wet-lab assembly (Beal, et al., ACS Synth Biol 1, 317-331, 2012; Slusarczyk, et al., Nat. Rev. Genet. 13, 406-420, 2012). As a result, researchers now have an ability to rationally design, model, and build networks to probe and control specific cellular behaviors, leading to potential therapeutic interventions and broader scientific discovery.

Synthetic biology’s techniques are performed with increasing robustness using a number of individual model species including E. coli. However, most naturally occurring bacteria live in communities, or consortia of multiple species (Nadell, et al., FEMS Microbiol. Rev.33, 206-224, 2009). Despite recent studies successfully demonstrating how engineered consortia can behave as a predator-prey systems or geospatially self-organize, the genetic network complexity required for targeted consortia engineering is daunting (Balagadde, et al., Mol. Syst. Biol. 4, 187, 2008; Brenner, et al., PLoS One 6, e16791, 2011; Mee, et al., Mol. Biosyst 8, 2470-2483, 2012). Additional modeling and engineering approaches are needed to further explore the details of microbiome interactions.

While examining the physiology of commensal microbes is important, macro-scale host behaviors must also be better understood to elucidate host-microbiome interactions. Biomimetic approaches give a robust toolset for analyzing animal behavior. For example, biomimetic robots have served as tools for exploring biomechanics ranging from snake locomotion to human balance (Ivanescu, et al., Bio-Inspired Models of Network, Information, and Computing Systems vol. 87, Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering (eds. Suzuki and Nakano) Ch. 54, 554-562, Springer Berlin Heidelberg, 2012; Luu, et al., IEEE Trans. Neural Syst. Rehabil. Eng. 19, 382-390, 2011). These robots provide quantifiable, minimal systems representative of natural phenomenal and are useful for scientific inquiry. In addition to mechanics, robots can be used to study cognition. By programming robots with a minimal set of algorithms and subjecting them to complex environmental challenges, researchers have used biomimetic robots to understand how primitive life forms solve a wide range of problems despite simple neural architectures (Ayers, et al., Philosophical transactions, Series A, Mathematical, physical, and engineering sciences 365, 273-295, 2007).

 

Model System: A Robotic Host with a Living Microbiome

In order to explore host-microbiome interactions, we created an in silico model system that combined the tools of synthetic biology and biomimetic robotics to design, model and computationally simulate a hybrid robot-bacteria system (FIGS.1A-1C and 2A-2F). First, a physical system that could be built was conceptualized, and contained a mobile robotic platform endowed with the capacity to harbor and communicate with a living microbiome. The designed system had three physical modules that could be built with existing technology (FIGS. 9A-9C). The individual modules were then modeled and simulated together to produce the complete system’s behavior. The final in silico model system was able to recapitulate a range of different biological‘host’ behaviors. Specifically, when the complexity of gene circuit topologies in the living microbiome was increased, unique robot behaviors were captured by this in silico tool. Model System: Physical Modules

The hybrid robot-microbiome system was designed to include three physical subsystems, or modules. These modules can exchange information through chemical, optical, and electrical signals. The first module (FIG. 9A) can be an engineered microbiome containing a living, synthetically engineered E. coli population. These bacteria can be engineered with gene circuits that drive fluorescent reporter expression (e.g., increases in green fluorescent protein (GFP) or mCherry - a red fluorescent protein - as shown in FIG. 2C). This engineered microbiome can be housed within the second module.

The second module can include a microfluidic chemostat²⁵ (FIG. 9B) that can be monitored by a miniature epifluorescent microscope²⁶ (FIGS. 2B and 9B). Cells can be well mixed and in exponential phase, similar to previous studies^19,27,28. Finally, this module would include electronics that sense and process the light signal from the epifluorescent microscope (EFM) into an electronic EFM signal (Table 1, FIG. 2C). This signal, hereafter referred to as the EFM signal, can be sent to the third module.

The third module can be the biomimetic robot host (FIGS. 2E and 9C). This third module can use the EFM signal to activate simple motion subroutines (Table 2, FIGS. 2F and 9C). The sum of these simple behaviors would then emerge as more complicated robot behaviors (FIGS. 1B, 1C and 2F). FIGS. 2A-2F illustrate how information would flow between the three modules. Here, the system is demonstrated with a simple synthetic circuit in the living microbiome, in which lactose drives GFP expression. All three modules are described in greater detail in Supplementary Text 1.

 

Model System: Computational Simulation of Proposed Physical Modules

Next, all three physical modules of the proposed robot-bacteria system were modeled and computationally simulated (FIGS. 10A-10C and Supplementary Text 1). The resulting complete system behavior was then simulated in a two-dimensional, virtual testing environment– or arena. This environment included stationary carbon source depots, which were conceptualized to contain inducers such as lactose or arabinose. These depots were conceptualized as prey for the robotic system. When the robotic system docked with these depots, it would capture the depot’s inducer carbon source and a potential fitness advantaged would be conferred to the host. The details of this‘docking’ are described in Supplementary Text 1.

The robot’s behavior in this simulated arena served as the output of our simulation. By analyzing the effects of variations in genetic circuit topologies and circuit parameter space, we observed a clear ability for the microbiome to cause distinct behavioral regimes in its host.

Results

Environment Simulation

In order to observe emergent robotic behavior, an environmental simulation scenario was designed. This scenario placed a robot in a 20 m x 20 m virtual, two-dimensional (2D) arena with an initial position at the center of this square. The simulated arena was initialized with one lactose and one arabinose carbon depot at different locations within the arena. These depots would remain in their position until the robot’s resulting movements led it to reach and dock at a depot, thereby acquiring all of the stored lactose or arabinose. This docking would cause the lactose or arabinose from the depot to enter the onboard microbiome at a constant concentration and rate. After a given carbon depot was depleted, a new source appeared of the same inducer (e.g., lactose or arabinose), but at a different location. In order for computational and analytical simplicity, the carbon sources were modeled to appear on the vertices of a 10 m x 10 m square, centered on the robot’s initial position. Finally, the biochemical environment of the microbiome was initialized with a simulated injection of arabinose at time t = 0. Simulation: Balanced Toggle Switching in Engineered Microbiome and Robot Behavior

In order to test the hypothesis that programming commensal bacteria with engineered gene circuits can result in new emergent host behavior, we different engineered gene circuits were simulated in the model system’s microbiome module. This was started by simulating a bistable memory element– or balanced toggle switch, based on the circuit initially developed by Gardner et al.¹⁰ due to the relative abundance of literature and  

characterization^29,30. This circuit allows confirmation that the system could capture the ‘toggle’ behavior of the microbiome circuity in the robot’s behavior, thus serving as a proof- of-concept of the designed system. The complete details of our analytical modeling and computational simulation approach is described in Supplementary Text 2.

The results of these simulations (FIGS. 3A-3E) show a robotic platform that alternates between seeking arabinose and lactose carbon depots. Upon initial activation with a transient pulse of an inducer, a balanced genetic toggle (FIG. 2A) drives sustained expression GFP or mCherry, and in the topology shown here, can be‘flipped’ by the external addition of either lactose or arabinose. The resulting temporal, biochemical landscape (FIG. 3D) drives a spatiotemporal robot behavior (FIG. 3E) characterized by its bistability. Repeated simulations always showed that a balanced toggle switch caused the robot to seek out a balanced set of depots (i.e., two lactose and two arabinose) in the virtual environment.

In order to probe this bistability further, the balanced toggle was simulated and included stochasticity in both transcription and translation^31,32 in the living microbiome. These simulations FIGS. 11A-11F showed a robotic platform that maintained its bistable memory, despite having additional stalls in motion. It should be noted that we found the robotic system behavior to be more sensitive to large degrees of stochasticity in translation, in comparison to transcription. Previous studies have attributed most stochastic variability in prokaryotic protein synthesis to noise in transcription³³ rather than translation, with the latter causing less overall noise due to transcriptional“bursts.”^31,34

Finally, the robotic platform’s toggling behavior at an environmental level was tested by simulating carbon depots that appeared randomly, rather than at fixed vertices (FIGS. 12A-12E). This simulation upheld the bistability of the robotic behavior, continuing to result in a robot that alternated its motion between lactose and arabinose sources regardless of the inducer carbon depot location.

Simulation: Biased Toggle Switching in Engineered Microbiome and Robot Behavior However, attaining a truly balanced toggle switch is difficult in the laboratory³⁵, and often wet-lab molecular bioengineering results in an unbalanced, or biased, toggle switch. This circuit lacks the stable equilibrium seen in the balanced toggle switch¹⁰, with a tendency to transcribe and translate one side of the circuit even when no inducers are present. This imbalance is driven by many factors including promoter strength, ribosome binding site (RBS) strength, and protein and mRNA degradation rates. A biased toggle switch can provide timer-like behavior useful for cellular control of processes ranging from metabolism to apoptosis. In order to discover how this genetic feature in the commensal microbiome alters host behavior, a biased toggle with an imbalance between the RBS strengths controlling LacI and TetR translation was simulated (See Supplementary Text 3).  

The resulting robot path (FIG. 4D) suggests a behavioral‘preference’ for lactose. This is attributed to the timer-like behavior of the genetic circuit demonstrated in the temporal reporter protein landscape (FIG 4C), wherein a spike in GFP production caused from exposure to lactose depots is quickly attenuated by a genetic bias for LacI and thus, mCherry synthesis. This behavior is caused by a difference in the LacI and TetR RBS ratio creating a translational imbalance for the repression proteins.

Simulation: Toggle Switch Parameter Sensitivity

The results shown in FIG. 4A-4D also raise an important question: without altering the genetic topology, what microbiome biochemical parameter(s) impact the emergent behavior of the robotic host? In order to explore this question, we performed a parameter sweep for the same RBS_LacI and RBS_TetR used to simulate FIG. 4A-4D. By evaluating quantitative metrics for behavior, such as simulation runtime and depots acquired, we were able to capture shifts in the robot’s behavioral regime (FIG. 5A-5D) driven exclusively by RBS strengths and the resulting toggle bias.

This parameter sweep also provided evidence of behavioral bifurcations seen in the yellow region to the upper right of FIG. 5C and in the intense black band in FIG. 5D. Although not immediately apparent, these areas represent distinctly different emergent robotic behavior than surrounding regions. The yellow region (FIG.5C) is characterized by a high percentage of time spent at zero velocity, which we defined as‘stalling’, for the robot and the black regions (FIG. 5D) represent time efficient host behavior, with a minimum amount of time spent stalling with an EFM value = 0. In conjunction, these differences in parameter space suggest potential host performance optimization caused by microbiome physiology.

Simulation: Host Feedback to the Engineered Microbiome

Having demonstrated different host behavior regimes that result from defined parameter spaces, we modified the genetic topology to include feedback from the robotic platform. This created a robotic simulation that included two way communication between the microbiome and the robotic platform. The resulting robotic behavior was more nuanced than the toggle switch, and was analogous to predation habits found in nature. By adding an orthogonal operon containing a P_lux-λ promoter³⁶ that simultaneously drives GFP and mCherry expression (FIG. 6A), we created a mechanism by which the microbiome can interpret the AHL pulse delivered by the host robotic platform (e.g., through the execution of subroutine 6 shown in Table 2). Thus, although it previously observed microbiome-to-host information flow, this additional circuitry allowed us to study host-to-microbiome feedback as well.

FIGS. 6A-6E details the results from simulating this additional engineered gene circuit within the microbiome. The simulation shows an interesting nuance in robot behavior.  

Rather than seeking out carbon depots directly, the robotic platform pauses, and then travels at twice the previous, base velocity as it nears the depot. This behavior - reminiscent of predation - is a natural analogy to what is known as a stalk-pause-strike³⁷ response in vertebrates.

Simulation: Feedback Parameter Sensitivity

Finally, the tunability in the host-microbiome feedback was demonstrated by adding an additional cI gene driven by the P_lux-λ promoter (FIG.7A) and modifying the corresponding RBS strength (RBScI). Our results show a number of distinct robotic behavioral regimes (FIGS.7A-7E) including both toggling and stalk-pause-strike behaviors previously noted.

The results from this RBScI parameter sweep capture four distinct robot behavioral regimes when varying RBScI from 0 to 1 (i.e., over the extremes of a relative range). When the RBScI value was close to zero (FIG. 7B), the stalk-pause-strike behavior seen in FIGS. 5A-5D was observed. This regime is expected as a low RBScI value (<0.0007) would imply that a negligible amount of cI is translated. As the RBScI value (0.0007-0.001) was increased, we observed a regime of stalk-pause-strike-pause-stalk previously unseen (FIG. 7C). Within this regime, the robot moved towards a depot (EFM ±1) and then stops motion (EFM = 0). After a brief pause, it began traveling at twice the base velocity (EFM ±2) before pausing again (EFM = 0) and finally finished its approach of the depot at base velocity (EFM ±1). Further raising the RBScI value (0.001-0.8) caused the robotic host to enter a regime of permanent stall (FIG. 7D), wherein the robot acquired no carbon sources. Finally, as the RBScI value approaches 1 (FIG. 7E), the robot behaves in the same bistable manner seen in FIGS. 3A-3E. This bistability is the result of quickly appearing, large [cI] that auto- represses its further transcription from the P_lux-λ promoter. These different regimes demonstrate the ability to tune robotic behavior by altering only a single genetic parameter within the microbiome.

Discussion

Although interconnectivity between commensal bacterial physiology and host behavior has been experimentally observed⁵, the underlying biochemical interactions³⁸ have yet to be fully understood. This Example provides an in silico tool that at least allows exploration of this relationship with synthetic biology. Much in the way that synthetic gene circuits allows the exploration of genetic pathways and relationships in a single organism³⁹, this system could be used to augment and examine the interconnected networks that drive host-microbiome interactions.

Here, two different topologies of information flow were explored that are important for host-microbiome interactions.

First, by simulating the toggle switch, information flow from the environment to the microbiome was explored, and then to the robotic platform. This system design (FIGS. 3A-  

3E) allowed establishment of an initial behavior theme: host alternation between nutrient sources (i.e., lactose and arabinose carbon depots) resulting from a repeatedly toggled, bistable gene network. It was then demonstrated that a translational parameter, RBS strength, could serve as a tunable component for modifying the robot’s affinity for these nutrient sources. Thus, it was possible to use both genetic topology and parameter strength to prescribe a range of robot behaviors (FIGS.4A-4D and 5A-5D).

However, host-microbiome systems in nature are not limited solely to microbiome-to- host communication. They also include mechanisms for host-to-microbiome information flow⁴⁰. By adding the additional P_lux-λ driven circuit and subroutine 6, we included this feature in our robotic system. In doing so, we simulated a system capable of mimicking host- microbiome interactions found in nature (FIGS.6A-6E). The addition of this circuit resulted in robot behavior analogous to stalk-pause-strike vertebrate predation³⁷. Furthermore, performing a one-dimensional parameter walk (i.e., varying the RBS strength driving cI expression) within this genetic topology showed that multiple distinct robot behaviors could be modulated by this single parameter (FIGS. 7A-7E). In addition to predation-like movement, these behaviors ranged from alternating between carbon source depots to permanently stalling. The results demonstrate that small changes in biochemical parameters can result in the emergence of very different host robotic behaviors.

The model system herein provides a useful system for exploring host-microbiome interactions with synthetic biology. By integrating an engineered microbiome, a microfluidic environmental niche, and a robotic conveyance, a biomimetic system has been designed, modeled, and simulated that allows us exploration of a natural phenomenon through both synthetic biological and robotic programming. This model system will have implications at least in the fields ranging from synthetic biology and ecology to mobile robotics.

Methods

All numerical simulations were programmed in MATLAB® 2014a using the Simulink™ software package. The simulations relied upon a combination of continuous and discrete functions, interacting in a block-based model, which is further described elsewhere herein. In order to facilitate accurate updating of state conditions, all integrations were calculated using MATLAB’s ode5 numerical method approach with a fixed time step. Ode5 is an implementation of the Dormand-Prince⁴¹ algorithm based off of Runge-Kutta approaches.

For every test cycle, all initial conditions were set at zero with the exception of internal arabinose concentration which had an initial condition of [50]. Each simulation ran through four iteration cycles corresponding to the four available carbon depots.

Data analysis and plotting of simulation results were performed using MATLAB and Python respectively. Python libraries numpy, scipy and matplotlib were leveraged for the

 

creation of graphics. Graphics were formatted as .SVG files and edited in InkScape® as vector images. Non-data graphics were created and edited exclusively in InkScape.

All Simulink and MATLAB files are made available by request from the authors.

All simulations were run on an ASUS Zenbook UX32VD running with an Intel® Core™ i7-3517U processor at 1.90 GHZ and 2.40 GHZ with 10 GB of RAM on a 64-bit Windows 8.1 operating system. Average runtime for each simulation trial was 2.78 minutes for the basic toggle switch (FIGS. 3A-3E) and incrementally more (<0.1 minute) for other circuits.

Supplementary Text 1: Robotic System Design

Module One: Microbiome

Module one describes a living, engineered microbiome. Microbiomes in nature are an amalgamation of numerous species (Tlaskalova-Hogenova, et al., Immunol. Lett.93, 97-108, 2004; Walter, et al., Annu. Rev. Microbiol.65, 411-429, 2011; Liu, et al., Genomics 100, 265- 270, 2012). Here a homogeneous, engineered E. coli population was used as a model system. This simplification is an established precedent and allows us to experiment and engineer with a well-understood model organism before adding layers of complexity (Brenner, et al., Trends Biotechnol.26, 483-489, 2008; Brenner, et al., PLoS One 6, e16791, 2011).

The effects of synthetic gene circuits within the microbiome were simulated (FIGS. 10A-10C) by using a deterministic approximation common in the synthetic biology (Gardner, et al., Nature 403, 339-342, 2000; Elowitz, et al., Nature 403, 335-338, 2000; Garcia-Ojalvo, et al., Proc. Natl. Acad. Sci. U.S.A. 101, 10955-10960, 2004). This framework centers on modeling the central dogma of molecular biology as a system of ordinary differential equations (ODE’s) relating the rates of change of inducers, mRNAs, and proteins. Although this approach neglects the stochastic variation found in nature, it provides a simplified, modular, and computationally efficient framework for understanding gene network dynamics (Elowitz, et al., Science 297, 1183-1186, 2002; Gillespie, et al., JCoPh 22, 403-434, 1976; MacDonald, et al., Integr. Biol. (Camb.) 3, 97-108, 2011; Wilkinson, et al., Nat. Rev. Genet. 10, 122-133, 2009; Tian, et al., Proc. Natl. Acad. Sci. U.S.A. 103, 8372-8377, 2006). This simplification appropriate for this research due to the homogeneous nature of the bacterial population (Rudge, et al, ACS Synth. Biol 1, 345-352, 2012). Nevertheless, the role of stochasticity in the system here was examined. Several simulations accounting for transcriptional and translational noise were created that are described elsewhere herein and shown in FIGS. 11A-11F (Raj, et al., Cell 135, 216-226, 2008; Ozbudak, et al., Nat. Genet. 31, 69-73, 2002).

The three primary equations for our ODE framework are shown below. An expanded derivation of these equations is presented in section Supplementary Text 2.  

Equation (Eq.) S1 is derived from analyzing the rate of change of the inducers inside of the cell. In the model described here, the endogenous network was simplified by envisioning a cell that had been engineered to no longer possess the native genes encoding arabinose and lactose metabolism; thus, the only incentive for seeking these inducers would be created by the engineered gene circuits. Therefore, the rate of change of the internal inducer concentration was approximated as a function of its transport across the cell membrane; this transport is governed by the gradient between the internal and external inducers concentration, [I_int] and [I_ex], respectively, and membrane permeability represented by a transport coefficient µ. In the model, inducers such as lactose and arabinose were introduced from the environment (FIGS. 3B, 4B, and 6B), or directly injected from the host robot (FIGS.6A and 7A). The transport processes leading to the first-principal derived model for inducer concentration is shown in FIGS. 10A-10C. Additional methods for modeling inducer concentration are noted in FIGS.10A-10C.

As noted, the model assumes a homogenous, well-mixed bacteria population, kept in the exponential phase of growth. These assumptions allowed for the description of the transcriptional and translational processes of synthetically engineered gene circuits as a mean population value, smoothing the fluctuations found at the single cell level (Karlebach, et al., Nat. Rev. Mol. Cell Biol. 9, 770-780, 2008). Equation S2 incorporates these assumptions to describe the rate of change of mRNA. Leveraging existing deterministic models, temporal dynamics of mRNA as the sum of four terms was modeled (Buse, et al., Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 81, 066206, 2010; Garcia-Ojalvo, et al., Proc. Natl. Acad. Sci. U.S.A. 101, 10955-10960, 2004). Under this paradigm, the first three terms on the right hand side of equation S2 relate the behavior of an inducible operon. Within these terms, [RP] is the concentration of a repressor protein with a corresponding Hill coefficient of H. α is a combined parameter describing a number of biophysical properties of the promoter site including transcription factor and RNAP binding affinity (Buse, et al., Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 81, 066206, 2010; Garcia-Ojalvo, et al., Proc. Natl.  

Acad. Sci. U.S.A. 101, 10955-10960, 2004; Vilar, et al., J. Cell Biol. 161, 471-476, 2003; Brewster, et al., E. Coli PLoS Comput Biol 8, e1002811, 2012). α_Leak is the rate of transcriptional leak of mRNA produced when the promoter site is repressed (Buse, et al., Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 81, 066206, 2010; Garcia-Ojalvo, et al., Proc. Natl. Acad. Sci. U.S.A. 101, 10955-10960, 2004). The inducer coefficient, k, is a parameter describing the rate at which mRNA is produced in proportion to the internal inducer concentration, [I_int]. Finally, the fourth term in the model describes kinetic rate of decay for the mRNA. Here a first order decay processes for the mRNA was assumed, represented by the HL_mRNA term. This transcription process is shown in FIGS.10A-10C.

Equation S3 describes the rate-of-change of protein within the cell. Specifically, this equation relates rate of protein produced with the concentration of mRNA within the cell. Fundamental to this model is the assumption that all mRNA transcribed can be translated. This assumption allows us to ignore mRNA inhibitors and riboregulators (Lopez, et al., Proc. Natl. Acad. Sci. U.S.A. 95, 6067-6072, 1998; Isaacs, et al., Nat. Biotechnol. 22, 841-847, 2004). Additionally, incorporation of a (ribosome binding site) RBS strength as a tunable parameter for altering system behavior was contemplated, which is an approach often used in synthetic biology (Cameron, et al., Nat. Rev. Microbiol. 12, 381-390, 2014; Mee, et al., Mol. Biosyst.8, 2470-2483, 2012; Esvelt, et al., Mol. Syst. Biol. 9, 641, 2013; Brophy, et al., Nat. Methods 11, 508-520, 2014; Kobayashi, et al., Proc. Natl. Acad. Sci. U.S.A.101, 8414- 8419, 2004). Therefore, a relative RBS strength term was included within equation S3. The protein’s rate of decay is approximated as first order, but of a different magnitude than that of the mRNA; this decay ratio is described by the parameter HL_ratio. This translation process is shown in FIGS.10A-10C.

Module Two: Microfluidic Chemostat

The design for an onboard, programmable microbiome leverages previous work in microfluidic based synthetic biology to approximate host-microbiome feedback found in nature (Huang, et al., Lab Chip 14, 3459-3474, 2014; Lee, et al., Lab Chip 11, 1730-1739, 2011; Groisman, et al., Nat. Methods 2, 685-689, 2005; Bennett, et al., Nat. Rev. Genet.10, 628-638, 2009). This module can be conceptualized as containing two features: 1) the physical chemostat (FIGS.9B and 2B) the miniaturized epifluorescent microscope (FIGS.2B and 9B).

The conceived chemostat (FIG. 9B) was based on existing designs and combines a microfluidic channel housing the first module with peristaltic pumps (Lee, et al., Lab Chip 11, 1730-1739, 2011; Bennett, et al., Nat. Rev. Genet.10, 628-638, 2009). By presuming that a carbon source, such as xylose, is constitutively pumped through the chemostat, we can assume the cells remain both well-mixed and in exponential phase (Lee, et al., Lab Chip 11, 1730-1739, 2011). Additionally, the microfluidic chip was conceptualized as a part of a   system permitting chemical injections from either the robot or the environment, allowing us to simulate a biomimetic proxy for information exchange with the microbiome.

The second module also contained a miniaturized epifluorescent (EFM) microscope based on previous designs (Ghosh, et al., Nat. Methods 8, 871-878, 2011). This module provided an interface translating phenotypic variations, in the form of reporter protein (mCherry and GFP) production (FIGS. 2B, 2C, 2D, and 9B), into electronic information encoded as voltage differentials. In this manner, the EFM serves as a biotic-abiotic interface.

In order to convert these reporter protein fluorescent measurements into a useful digital signal, a computable response function (Table 1) to interpret the reporter protein intensities and to translate the measurement into one of five discrete outputs was designed. This logic function generates what was termed the EFM value, a signal sent to the robotic host microprocessor. EFM signal value thresholds were set to allow responses to distinct regimes of reporter protein concentrations outputted by our simulated model. By extension, in a physical system, these values can be based on fluorescent intensity recorded by the epifluorescent microscope.

Table 1

Module Three: Robotic Mobile Prosthetic and Microprocessor

The third module is a robotic host and a microprocessor that controls all mechatronic behavior for the robotic platform. In our simulation, the robot was designed to have mobile functionality similar to the e-puck swarm robot (Cianci, et al., Swarm Robotics Vol. 4433, Lecture Notes in Computer Science, Ch. 7, 103-115, Springer Berlin Heidelberg, 2007). Therefore, the robot was a tank robot with two wheel actuation (Braitenberg, V. Vehicles: Experiments in Synthetic Psychology, 1E, MIT UP, 1984). We designed the robot to have external sensors capable of locating and differentiating different inducer (lactose or arabinose) carbon depots.

The robotic host also includes hardware that allows it to‘dock’ with an inducer carbon depot. This hardware would establish a watertight connection between the mobile robotic platform and the carbon depot. Once this seal has been established, the docking port would   allow for the inducer to enter the microchemostat at a constant flow rate. During this docking, the robotic platform is still sensitive to the signals sent from the EFM.

Furthermore, the robot was programmed with a minimal set of subroutines (Table 2) designed to mimic an organism’s mobile pursuit of nutrients (e.g., hunting) within its environment (FIGS. 1A-1C). These commands were programmed into an onboard microprocessor that reevaluates subroutines states at every time step of the simulation (FIGS. 9A-9C). This minimal set of subroutines allowed us to observe how the phenotypic state of the microbiome influences host behavioral response.

Table 2: Programmed Robot Subroutines

The first five subroutines relate the robot’s motion within the simulation environment. The last subroutine injects a pulse of a third inducer into the microchemostat directly from the robotic platform. This subroutine, sixth in Table 2, is a simplification of the host-to- microbiome biochemical communication interaction found in nature. By including this biomimetic feature, we were able to create information exchange from the host to the microbiome (FIGS. 13A-13E), in addition to the other subroutines that enabled microbiome information to be passed to the host. It was chosen to mimic and simulate this biochemical communication using N-Acyl homoserine lactone (AHL) as the inducer molecule along with an AHL-sensitive, engineered promoter, P_lux-λ, due to this system’s orthogonally and well characterized behavior^8,28,36-38.

  Supplementary Text 2 (Text S2): Model Derivation

Robot System Design

The results in this Example are based on our design for a biomimetic robotic platform, engineered to simplify the host-microbiome interactions found in nature. This system allows capture of five crucial information flows: environment-to-host (external sensors), environment-to-microbiome (arabinose and lactose), host-to-microbiome (AHL pulse), microbiome-to-host (epifluorescent signal output) and host-to-environment (robot position). This information flow is presented at least in FIGS.1A-1C and FIG.9A-9C.

Biochemical Model Development

The biochemical simulation used in our model links inducer, mRNA, and protein concentrations. The field of computational molecular biophysics is vast, with established models existing for different scales and system complexity. One of the nuances when modeling cellular behavior is to select a modeling approach that captures the critical behavior while providing computational efficiency. For instance, one would not use a molecular dynamics model to explain the human circulatory system.

For the studies presented in this Example, the model reporter protein expression resulting from inducer concentrations that activate synthetically inserted genetic topologies was modeled. Within this context, synthetically programmable parameters, such as promoter and RBS strengths, were related in order to test our hypothesis. Fortunately, there is no shortage of existing models for understanding the transcription and translation processes. For example, the rise of systems and synthetic biology in conjunction with biophysics and numerical methods has opened the door for Monte Carlo/Markov Chain (MMMC) models used for even complex genomic networks (Gillespie, et al., JCoPh 22, 403-434, 1976).

However, although an understanding of the stochastic nature of gene expression is important, the design of the system described herein allowed use of a continuous, deterministic approximation. This is a fair assumption given that the system focused on macroscopic characteristics (reporter protein fluorescence) at a population level for a homogeneous culture.

Using a continuous framework, the first task was to describe the inducer concentration within the cell. From an inducer specific mass balance, we can describe change in inducer concentration as the sum of the transport of the inducer across the cell membrane and the degradation of the inducer by cellular kinetics.

  Further exploring the terms on the right hand side, we can describe inducer transport into the cell by modifying Fick’s law for diffusion over a transport barrier, given in equation S5.

In this equation, J is the diffusion flux measured in units of concentration per unit area per unit time, dc/dt is the inducer concentration gradient, and ^D is a transport diffusion coefficient that describes the ability for substances to flow through the membrane.

The reactor was designed to be well mixed, and therefore dc/dt can be assumed to be equal at all locations across the cell membrane for a given moment in time. This assumption allowed the assumption that equation S3 is true.

Whereby is the inducer concentration external to the cell and is the inducer

concentration within the cell membrane.

Additionally, within cells, D would normally behave as a function of membrane channel proteins such as AraFGH and permease^39,40. However, for model simplification, it was assumed these membrane proteins were held at a constant concentration, and therefore D was a constant. By additionally assuming all cells in the microbiome have a constant surface area, Fick’s law can be modified by the assumptions to yield.

Within equation S4, µ is a combined transport coefficient, defined as µ = D×A, with A being the cell’s average surface area.

The second term on the right side of equation S1, inducer metabolism, relies on a kinetic model to characterize the degradation of the inducer once inside the cell membrane. However, to simplify metabolism kinetics, it was assumed that the engineered E. coli would include gene knockouts that eliminate the metabolism of the inducers used in our system; lactose, arabinose, and AHL. Therefore, the internal concentration of these three inducers may be characterized by equation S5.

It should be noted that for all three of these inducers, [I_ex] are controlled by non- cellular factors such as proximity to carbon depots and robot executable 6. In this way, the inducers serve as external signals linking information flow from the environment or robotic platform to the microbiome. It is noted that equation S8 represents first-order kinetics resulting in exponential decay of inducer concentration.   For the next variable, mRNA, a mass balance was first set up, noting that the rate of change for mRNA will equal the transcription rate minus the degradation rate. This equation S9 is formulated similarly to S4:

In order to develop a term for the rate of transcription, we turned to existing literature for model development. However, it was desired to make a modular, continuous approximation of operon behavior and therefore did not want to engage in promoter-site specific Shea-Acker’s formalism⁴¹. Existing work in synthetic genetic network dynamics has been explored^7,8,15 has developed a simplified continuous ODE for approximating inducible and repressible operon for behavior.

This equation relates the rate of change in mRNA production to a number of inputs driven by transcription events and mRNA degradation.

The first term on the right hand side is used to describe how the concentration of repression proteins [RP], such as TetR or LacI, affect the normal promoter-driven gene expression. Within this term, α^is a coefficient describing the maximum transcription rate when no repression proteins are present. Finally, H is a term known as the Hill coefficient, and is used to describe the relative impact of a repression protein on an associated promoter.

The second term, α_Leak, is a term describing the‘leak’ of a promoter. This term varies in accordance to the promoter studied. However, in our simulation we kept α_Leak to be roughly 1/100 of the α value.

The final term provides a mechanism for induced operon activation. Within this term, k is a signal coefficient that relates the amount of inducer to the rate of transcription. Finally, for the rate of mRNA degradation, we developed a simple kinetic model relating the degradation rate to the half-life and concentration of the mRNA.

It should be noted that many biological factors can alter the mRNA half-life such as nuclease tags. Furthermore, an mRNA’s translational efficacy may be altered by the

  presence of riboregulators and other inhibitors. However, assuming these factors are not present, equations S9, S10, and S11 can be combined.

Using a similar, but simpler model for protein translation and degradation, it was possible to model the rate of change for protein as equation S10.

Where [P] is the protein concentration, RBS is the ribosome binding site strength associated with the mRNA strand, and HL_ratio is the half-life ration of the mRNA to protein half-lives. With S8, S12, and S13, a set of governing biochemical equations shown in Supplementary Text 1 was able to be written. Governing Biochemical Equations

These governing equations are also shown in FIGS.10A-10C. Supplementary Text 3 (Text S3): Simulation Design

In order to simulate the interplay between host, microbiome, and environment, we designed a block system architecture that combined continuous biochemical readouts with finite-state-machine logic functions for the epifluorescent microscope and microprocessor state. The general concept for this design is presented below in FIGS.12A-12E.

Within the simulation, a fixed time step evaluation of state variables (protein, mRNA, inducer, position) was used to determine which subroutine the robotic platform should run. With this fixed time step evaluation in mind, a fixed time step numerical method to evaluate the biochemical system was selected. In this way it was able to best represent the engineering constraints of our designed biomimetic robot.   All of the Simulink and MATLAB files detailing the model parameters are available by request from the authors.

Balanced and Biased Toggle Switch

Here, the set of governing ODEs for the genetic toggle switch presented in this Example are included:

The simulation had the following parameter values unless otherwise noted: HL_{mRNA_Pbad ,}HL_{mRNA_Plac,} HL_{mRNA_Ptet} = 1; α_Pbad = 0; α_Ptet, α_Plac = 0; α_Pbad = 200; HL_{ratio_mCherry,} HL_{ratio_GFP,} HL_{ratio_TetR} HL_{ratio_LacI} = 3; RBS_GFP, RBS_mCherry, RBS_etR, RBS_LacI= 1; α_{Leak_Ptet,} α_{Leak_Plac,} α_{Leak_Pbad} = 1; k_pbad, k_plac = 50; k_ptet = 0; Hill_lac, Hill_tet = 2. These parameters are based off of a calculation regime used in previous literature⁸. Additionally, Lactose_DoseConc, Arabinose_DoseConc = [50].

FIGS. 4A-4D was generated by changing the RBS_LacI equal to 2.4. This value was chosen as a visually indicative change in the behavioral regime.

FIGS. 5A-5D is the result of a two dimensional parameter sweep changing the RBS_TetR and RBS_LacI values from 1-10 by increments of 1. The simulations were run in series and the quantitative metrics describing the behavior for each RBS combination was assembled in an array. The array is visually represented by the heat contours shown in FIGS.5A-5D.

  Network Stochasticity for the Toggle Switch

First, inducer stochasticity was simulated by incorporating a Gaussian kernel as a multiplier of the exponentially decaying internal inducer concentration. The results from this simulation are presented in FIGS. 14A-14D. Due to the toggle switch’s bistable nature, relatively small amounts of inducer stochasticity had negligible effects on the robotic emergent behavior (Tian, et al., Proceedings of the National Academy of Sciences 103, 8372-8377, 2006).

To account for gene expression stochasticity, the governing biochemical equations were modified to include noise terms. These additions were based upon examples from literature that augment the ODE’s to include the stochastic terms ηR and ηP for transcriptional and translational noise, respectively (Ozbudak, et al., Nat. Genet.31, 69-73, 2002).

In the continuous example, the terms ηR and ηP are defined as being random variables with a normal (Gaussian) distribution about a mean of zero. Within the model, the variance for ηR and ηP is defined as percentage of the mRNA or protein concentration at a given time step. For instance, for a given percentage, ρ, ηR and ηP are defined by S14 and S15 below, where is a normal distribution as a function of the mean (ૄ) and the variance (ૅ).

The random variables were then incorporated into the equations for translation S2 and transcription S3 to arrive at the stochastic versions of transcription and translation shown in equations S16 and S17, respectively.

ρ = {0%, 1%, and 5%} was simulated for the transcription and translation of all mRNA and protein products. The results are shown in FIGS.11A-11F and FIGS.15A-20D.

Finally, in to consider environmental stochasticity the lactose and arabinose depots were simulated to appear at random locations. The results for four trial runs are shown in FIGS. 12A-12E.

Additional P_lux-λ Operon

The simulations including the P_lux-λ operon (FIGS. 6A-6E and 7A-7E) used a slightly different set of equations to accommodate the additional behavior.  

Furthermore, the additional circuit added the following modeling parameters: AHL_near = 2.0; AHL_far = 2.25; AHL_DoseConc = [130]; HL_{mRNA_PLux-λ} = 1 α_Plux-λ = 0; α_{Leak_Plux-λ} = 1; k_PLux-λ = 100; HL_{ratio_cI} = 1; RBS_{GFP_2}, RBS_{mCherry_2} = 4.

For FIGS.7A-7E, the RBScI value was varied from 0 to 1 and it was observed where regime shifts occurred. The simulation results were visually inspected and areas of regime bifurcations were found. The selected regimes are indicative of the major visually observed behavioral shifts. References for Example 1

1 Backhed, F. et al. The gut microbiota as an environmental factor that regulates fat storage. Proc. Natl. Acad. Sci. U. S. A.101, 15718-15723,

doi:10.1073/pnas.0407076101 (2004).  

2 Markle, J. G. M. et al. Sex Differences in the Gut Microbiome Drive Hormone- Dependent Regulation of Autoimmunity. Science 339, 1084-1088,

doi:10.1126/science.1233521 (2013).

3 Tlaskalova-Hogenova, H. et al. Commensal bacteria (normal microflora), mucosal immunity and chronic inflammatory and autoimmune diseases. Immunol. Lett.93, 97- 108, doi:10.1016/j.imlet.2004.02.005 (2004).

4 Sharon, G. et al. Commensal bacteria play a role in mating preference of Drosophila melanogaster. Proc. Natl. Acad. Sci. U. S. A.107, 20051-20056,

doi:10.1073/pnas.1009906107 (2010).

5 Neufeld, K. M., Kang, N., Bienenstock, J. & Foster, J. A. Reduced anxiety-like

behavior and central neurochemical change in germ-free mice. Neurogastroenterol. Motil.23, 255-264, e119, doi:10.1111/j.1365-2982.2010.01620.x (2011).

6 Hughes, D. T. & Sperandio, V. Inter-kingdom signalling: communication between bacteria and their hosts. Nat Rev Micro 6, 111-120 (2008).

7 Walter, J. & Ley, R. The human gut microbiome: ecology and recent evolutionary changes. Annu. Rev. Microbiol.65, 411-429, doi:10.1146/annurev-micro-090110- 102830 (2011).

8 Greenblum, S., Chiu, H. C., Levy, R., Carr, R. & Borenstein, E. Towards a predictive systems-level model of the human microbiome: progress, challenges, and

opportunities. Curr. Opin. Biotechnol.24, 810-820, doi:10.1016/j.copbio.2013.04.001 (2013).

9 Brophy, J. A. & Voigt, C. A. Principles of genetic circuit design. Nat. Methods 11, 508-520, doi:10.1038/nmeth.2926 (2014).

10 Gardner, T. S., Cantor, C. R. & Collins, J. J. Construction of a genetic toggle switch in Escherichia coli. Nature 403, 339-342, doi:10.1038/35002131 (2000).

11 Elowitz, M. B. & Leibler, S. A synthetic oscillatory network of transcriptional

regulators. Nature 403, 335-338, doi:10.1038/35002125 (2000).

12 Anderson, J. C., Voigt, C. A. & Arkin, A. P. Environmental signal integration by a modular AND gate. Mol Syst Biol 3, 133, doi:10.1038/msb4100173 (2007).

13 Ellis, T., Wang, X. & Collins, J. J. Diversity-based, model-guided construction of synthetic gene networks with predicted functions. Nat. Biotechnol.27, 465-471, doi:10.1038/nbt.1536 (2009).

14 Friedland, A. E. et al. Synthetic gene networks that count. Science 324, 1199-1202, doi:324/5931/1199 [pii]10.1126/science.1172005 (2009). 15 Daniel, R., Rubens, J. R., Sarpeshkar, R. & Lu, T. K. Synthetic analog computation in living cells. Nature 497, 619-623, doi:10.1038/nature12148 (2013).

 

16 Beal, J. et al. An end-to-end workflow for engineering of biological networks from high-level specifications. ACS Synth Biol 1, 317-331, doi:10.1021/sb300030d (2012). 17 Slusarczyk, A. L., Lin, A. & Weiss, R. Foundations for the design and implementation of synthetic genetic circuits. Nat. Rev. Genet.13, 406-420, doi:10.1038/nrg3227 (2012).

18 Nadell, C. D., Xavier, J. B. & Foster, K. R. The sociobiology of biofilms. FEMS

Microbiol. Rev.33, 206-224, doi:10.1111/j.1574-6976.2008.00150.x (2009).

19 Balagadde, F. K. et al. A synthetic Escherichia coli predator-prey ecosystem. Mol.

Syst. Biol.4, 187, doi:10.1038/msb.2008.24 (2008).

20 Brenner, K. & Arnold, F. H. Self-organization, layered structure, and aggregation enhance persistence of a synthetic biofilm consortium. PLoS One 6, e16791, doi:10.1371/journal.pone.0016791 (2011).

21 Mee, M. T. & Wang, H. H. Engineering ecosystems and synthetic ecologies. Mol Biosyst 8, 2470-2483, doi:10.1039/c2mb25133g (2012).

22 Ivanescu, M., Bizdoaca, N., Hamdan, H., Eltabach, M. & Florescu, M. in Bio-Inspired Models of Network, Information, and Computing Systems Vol.87 Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering (eds Junichi Suzuki & Tadashi Nakano) Ch.54, 554-562 (Springer Berlin Heidelberg, 2012).

23 Luu, B. L., Huryn, T. P., Van der Loos, H. F., Croft, E. A. & Blouin, J. S. Validation of a robotic balance system for investigations in the control of human standing balance. IEEE Trans. Neural Syst. Rehabil. Eng.19, 382-390,

doi:10.1109/TNSRE.2011.2140332 (2011).

24 Ayers, J. & Witting, J. Biomimetic approaches to the control of underwater walking machines. Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 365, 273-295, doi:10.1098/rsta.2006.1910 (2007).

25 Bennett, M. R. & Hasty, J. Microfluidic devices for measuring gene network dynamics in single cells. Nat. Rev. Genet.10, 628-638, doi:10.1038/nrg2625 (2009).

26 Ghosh, K. K. et al. Miniaturized integration of a fluorescence microscope. Nat.

Methods 8, 871-878, doi:10.1038/nmeth.1694 (2011).

27 Lee, K. S., Boccazzi, P., Sinskey, A. J. & Ram, R. J. Microfluidic chemostat and turbidostat with flow rate, oxygen, and temperature control for dynamic continuous culture. Lab Chip 11, 1730-1739, doi:10.1039/c1lc20019d (2011).

28 Groisman, A. et al. A microfluidic chemostat for experiments with bacterial and yeast cells. Nat. Methods 2, 685-689, doi:10.1038/nmeth784 (2005).

 

29 Kobayashi, H. et al. Programmable cells: interfacing natural and engineered gene networks. Proc. Natl. Acad. Sci. U. S. A.101, 8414-8419,

doi:10.1073/pnas.0402940101 (2004).

30 Tian, T. & Burrage, K. Stochastic models for regulatory networks of the genetic

toggle switch. Proc. Natl. Acad. Sci. U. S. A.103, 8372-8377,

doi:10.1073/pnas.0507818103 (2006).

31 Raj, A. & van Oudenaarden, A. Nature, Nurture, or Chance: Stochastic Gene

Expression and Its Consequences. Cell 135, 216-226, (2008).

32 Elowitz, M. B., Levine, A. J., Siggia, E. D. & Swain, P. S. Stochastic gene expression in a single cell. Science 297, 1183-1186, doi:10.1126/science.1070919 (2002). 33 Ozbudak, E. M., Thattai, M., Kurtser, I., Grossman, A. D. & van Oudenaarden, A.

Regulation of noise in the expression of a single gene. Nat. Genet.31, 69-73 (2002). 34 Thattai, M. & van Oudenaarden, A. Intrinsic noise in gene regulatory networks.

Proceedings of the National Academy of Sciences 98, 8614-8619,

doi:10.1073/pnas.151588598 (2001).

35 Khalil, A. S. & Collins, J. J. Synthetic biology: applications come of age. Nat. Rev.

Genet.11, 367-379, doi:10.1038/nrg2775 (2010).

36 Tabor, J. J. et al. A synthetic genetic edge detection program. Cell 137, 1272-1281, doi:10.1016/j.cell.2009.04.048 (2009).

37 Moermond, T. C. Prey-attack Behavior oiAnolis Lizards. Z. Tierpsychol.56, 128-136, doi:10.1111/j.1439-0310.1981.tb01291.x (2010).

38 Lyte, M. Microbial endocrinology in the microbiome-gut-brain axis: how bacterial production and utilization of neurochemicals influence behavior. PLoS Pathog.9, e1003726, doi:10.1371/journal.ppat.1003726 (2013).

39 Ruder, W. C., Lu, T. & Collins, J. J. Synthetic biology moving into the clinic. Science 333, 1248-1252, doi:10.1126/science.1206843 (2011).

40 Montiel-Castro, A. J., Gonzalez-Cervantes, R. M., Bravo-Ruiseco, G. & Pacheco- Lopez, G. The microbiota-gut-brain axis: neurobehavioral correlates, health and sociality. Front. Integr. Neurosci.7, 70, doi:10.3389/fnint.2013.00070 (2013).

41 Dormand, J. R. & Prince, P. J. A family of embedded Runge-Kutta formulae. JCoAM 6, 19-26, doi:10.1016/0771-050x(80)90013-3 (1980).

 

Example 2: Introduction: Over the past twenty years, significant advances in robot decision- making architectures have been made, ranging from onboard deep neural networks to multiple-agent shared planning (LeCun, et al., Nature 521, 436-444; Howard, et al., The International Journal of Robotics Research 25, 1243-1256). Many of these systems seek to replicate the advanced information processing exhibited by human and other animal brains. From this perspective, biological inspiration is a significant driver of advances in robotics. Biologically inspired robots have also been useful in understanding scientific phenomena. Many of these studies have explored complex kinematics by building biomimetic robots that recapitulate the movements of diverse species ranging from hummingbirds and insects to lizards and snakes (Matthew, et al., Development of the Nano Hummingbird: A Tailless Flapping Wing Micro Air Vehicle, in 50^th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, American Institute of Aeronautics and Astronautics, 2012; Shang, et al., Bioinspiration and Biometrics 4, 2009; Feifei, et al., Bioinspiration and Biometrics 10, 2015; Marvi, et al., Science 346, 224-229, 2014). These types of kinematically inspired robots have diverse applications, ranging from underwater mine detection to minimally invasive surgery (Button, et al., A survey of missions for unmanned undersea vehicles, DTIC Document, 2009; Simaan, et al., A dexterous system for laryngeal surgery, in Robotics and Automation, 2004. Proceedings. ICRA’04. 2004 IEEE International Conference on, pp.351-357, IEEE.

Researchers have also used biomimetic systems to explore neurophysiology and understand the organization of the body’s central and peripheral nervous systems. By mimicking neurological architectures with simple in silico heuristics, engineers have created highly adaptable robots with minimal computational burden (Steingrube, et al., Nat Phys 6, 224-230, 2010). Examples include advanced submerged and swarm robots (Ayers, et al., Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 365, 273-295, Tan, et al., Defense Technology 9, 18-39, 2013). Similar attempts to emulate segments of living brains in silico have enhanced scientists’ understanding of human and animal cognition and sparked breakthroughs in problem- solving artificial intelligence (Markram, et al, Cell 163, 456-492; Silver, et al., Nature 529, 484-489, 2016).

Yet, biologists are realizing that these systems are not the only significant processing hubs in the body. Microbiologists and systems biologists increasingly connect information processing by the microbiome– the collection of all of the microbes living on and in the body – with animal physiology and decision-making (Montiel-Castro, et al., Frontiers in the Integrative Neuroscience 7, 2013). For example, commensal microbes, such as the bacteria  

in the gastrointestinal tract, have been shown to affect emotions such as anxiety and to correlate with behaviors such as reproductive affinity (Bercik, et al., Gastroenterology 141, 599-609; Sharon, et al., Proceedings of the National Academy of Sciences 107, 20051- 20056). This connection between information processing by the gut and information processing by the brain is frequently referred to as the“gut-brain axis” (Montiel-Castro, et al., Frontiers in Integrative Neuroscience 7, 2013). Previously, we developed a basic analytical and computational model of this interaction that illustrated how a biomimetic system could be used to explore biological phenomena (Heyde, et al., Scientific Reports 5, 11988). In the work presented here, we go beyond basic biological exploration to develop new heuristics and control schemes for robot behavior that mimic the parallel information processing of the microbiome in conjunction the body’s nervous system.

This Example demonstrates the development of a new paradigm for robot control. In doing so, a robot was engineered with two interconnected, information-processing hubs that cooperatively dictate robot decision-making. One hub mimicked the microbiome by simulating a population of engineered bacteria cells. The other hub, designed to mimic a host organism, consisted of a physical robot programmed with a set of basic actuating subroutines. This system design allowed development of robot control structures composed of hybrid systems containing both continuous actors, such as simulated engineered living cells, and discrete, programmed state machines within the robot’s embedded electronics.

Materials and Methods

The biomimetic gut-brain axis system provided herein was composed of two information- processing hubs: one hub was a physical robot with an embedded microcontroller and the other hub was a simulated population of engineered cells. These hubs were linked by a finite state machine that relayed conditions via a Bluetooth serial port. This system architecture allowed us to encode different relationships between the host and microbiome processing hubs, enabling us to solve problems with novel models of computation composed of both biological simulations and conventional embedded processors.

Physical Host Robot

A simple differential drive robot chassis was built (FIG. 21A) with two actionable wheels (degree of: steerability = 0, mobility = 2, maneuverability =2) and two caster wheels for stability. Despite being non-holonomic, this design was chosen due to its canonical nature in biomimetic design¹⁸ and ease of analysis. The simple and well-established kinematics of this system allowed us to focus on how the interplay between our two information-processing hubs affect robot performance, without concerning ourselves with nonconventional feedback control.

An open source Arduino® Leonardo (Sparkfun Cat. No. Dev 11286), with integrated motor drivers, was used as the onboard, embedded microcontroller. This system was  

chosen due to its simplicity, reliability, and ease of add-ons open source breakouts and libraries readily available. The firmware was programmed entirely in the Arduino language (C/C++ subset) and uploaded to the robot platform prior to each set of experiments.

By using a camera with integrated computer vision (Charmed Labs^® CMUcam5), the Arduino microcontroller could directly analyze images onboard. This camera enabled the robot to recognize objects in the testing arena and estimate distance based on the relative image size.

Finally, the mobile robot interfaced with the second information-processing hub, a simulated population of engineered cells, via a Bluetooth relay. By connecting the microcontroller to a DFRobot^® Serial Bluetooth Module (Cat. No. RB-Dfr-10), a serial channel between the mobile robot and a finite state machine actor was opened that relayed the condition of the microbiome processing hub. This allowed for easy information exchange between the simulated cell actor and the robot’s embedded processor.

Simulated Microbiome Population

The second information hub was composed on a simulated biomimetic microbiome, modeled with a synthetically engineered population of E. coli bacteria cells. This information hub was programmed to behave as a continuous actor, with small chemical inducers of bacterial gene networks, such as lactose, arabinose or N-acyl homoserine lactone (AHL), mimicked by input ports and signal proteins, such as fluorescent proteins, mimicked as output ports.

The microbiome model was developed from first-principle chemical kinetics, employing a Michaelis-Menten formalism¹⁹ in line with previous computational^{17, 20} and experimental^{21, 22} studies by ourselves and others. Although stochasticity can affect cellular chemical reactions, we concluded that a deterministic model could sufficiently approximate the temporal dynamics of cellular biochemical reactions in our system^{17, 23}. A presentation and motivation for the governing equations are presented in the supplementary information as well as previous literature¹⁷.

The synthetic gene regulatory networks were designed to contain sequences encoding for fluorescent reporter proteins such as green fluorescent protein (GFP) or mCherry, a red fluorescent protein. In practice, these reporter proteins cause a detectable phenotypic variation that can be monitored by an epifluorescent light microscope. However, for the purpose of developing host-microbiome inspired decision architectures, GFP and mCherry were considered as two cases of a generic class of reporter proteins that serve as signal outputs from microbiome actor.

Cell-Robot Interface

The microbiome hub was connected to the mobile physical robot by a MATLAB® relay program. This piece of software mimicked the functionality of a miniature epifluoresent  

microscope²⁴ while allowing our simulated cells to interface with a moving robot. First, the relay measures the signal protein concentrations (FIG. 21C) within the simulated bacteria population. Predetermined thresholds then map this continuous signal onto a discrete set of integers. These integer outputs were named as the digital epiflouresecent microscope (EFM) values (FIG. 21D). The EFM values are then transmitted to the physical robot via Bluetooth serial connection. Once onboard, the signals are processed into actionable, mechatronic, subroutines, such as seeking a specific target.

Testing Arena

In order to quantify the robot’s behavior under different decision architectures, a controlled testing environmental, or arena, was built. A5 m x 5 m, two-dimensional (2D) space in a pilot laboratory was cleared and taped. The arena contained the robot, yellow taped markers, and green or red targets. When the robot was in contact with one of these targets, a chemical input would be delivered to microbiome processing hub at a constant rate and concentration.

Within our system, information flows (FIG. 21E) from the simulated cell to the robot via Bluetooth in the form of an EFM value. The microcontroller then converts this EFM value into a specific subroutine controlling robot movement within a testing environment. This testing arena contained physical objects, such as different colored cylinders, that served as targets for the robot. Information from the environment is transmitted to the simulated cell by chemical signals corresponding to the robot’s contact with a target. Taken together, the information exchange between the simulated cell, cell-robot interface, and physical robot allows us to explore how we can encoded different relationships between the host and microbiome to solve problems.

Results

Shared Decision Architectures

Using the biomimetic robot, an objective was to develop robust control architectures for locating a simple target within an arena. A simple heuristic found in nature is the“run and tumble” chemotaxis motif^{25, 26}, in which periods of pure translation are followed by periods of pure rotation. In practice, this behavior allows simple organisms to navigate nuanced environments in a manner akin to a biased random walk. Despite its subtlety, this behavior is robust and energetically optimized²⁷.

Efforts were made to engineer different relationships, between the host and microbiome processing hubs, which could cause the robot to converge to a target while employing a biased random walk. To explore the efficacy of the system we a testing scenario in which a robot would be placed two meters away from a green target was developed. Once the biological simulation was started, the robot’s motion would be tracked (FIG. 22A) until the robot touched the target or exited the arena.The motion path found in  

FIG.22A was taken from a video recording of the robot’s behavior. This simple experimental design allowed us to compare results stemming from three hybrid architectures: master/slave, run and tumble, and integrated perception.

Master/Slave Architecture

The master/slave architecture was the simplest method for integrating engineered cells with the robot’s control structure. By linking elevated levels of signal proteins with robot subroutines initiating movement towards different targets, such as the green or red targets, the microbiome hub could issue“master” commands (FIG.22B) to the“slave” robot.

The microbiome processing hub was modeled to contain a bistable, mutually repressible gene circuit (FIG.22C). This regulatory gene network, known as a toggle switch, is common in synthetic biology literature²² and has been modeled and experimentally validated many times. The gene network contains a primitive form of biochemical memory. When a given chemical input, such as lactose, enters the cell, a sustained production of a repressor protein attenuates one side of the gene network. Thus, there is a bistability inherent in the gene network. The sustained expression of a given signal protein is relayed to the robot, causing a biased random walk towards the green target (FIG.22A).

This strategy for hybrid control architecture is simple and requires minimal integration of the microbiome processing hub within the robot’s host processing hub. Although this decision architecture is hybrid, it relies almost exclusively on a top down flow of commands from the microbiome processing unit to the embedded host processing hub.

Run and Tumble Architecture

In the master/slave architecture, the robot performs two distinct locomotive phases of translating and rotating when executing a biased random walk. The differential drive robot chassis can execute these behaviors by simply inverting the angular velocity of one of the two fixed wheels. It should be noted that for simplicity, no corrective feedback control was added to these locomotion primitives.

By modifying the control structure of both the microbiome and host processing hubs, we can reproduce the biased random walk with a new decision architecture. These results (FIG. 23A) show a robot moving through an arena, seeking the green target, with distinct periods of translation (green dash) and rotation (red arrow) locomotion. The motion path found in FIG.23A was taken from a video recording of the robot’s behavior. These behaviors were the result of a control architecture (FIG. 23B) that allowed relative levels of signal proteins to control the angular velocity of the fixed wheels. This has the effect of linking periods of rotation and translation to signal protein levels.

Chemical pulses were delivered to the microbiome processing hub after a rotation subroutine or when the green target increased in size. These chemical pulses served as information packets, transmitting environmental data from the robot to the simulated cells.  

Much in the way a host organism can release hormones that affect the microbiome, this chemical pulse allowed our host processing hub to affect the biochemical simulation of the microbiome processing hub.

Cells within the microbiome processing hub contained a regulatory gene network (FIG.23C) similar to the mutually repressible network used in the master/slave architecture, but lacking the bistability. Whereas the master/slave gene network allowed for long-term memory, network used here contained a bias towards one side, causing timer-like behavior^{28, 29}. This bias was caused by elevated level of repression protein production, a feature encoded within the genes themselves. This behavior can be observed in the graph of the signal proteins (FIG.23D), which shows a time-linked attenuation of signal protein 1.

By exploiting the timer-like nature of the biased toggle, the simulated cells drive the robot by calling for alternating periods of translation and rotation. Only when there is an increase in the target size is additional lactose introduced to the cells. This feature results in periods of sustained forward translation, as noted at simulation time 260 h. The net effect is a biasing of the robot’s random walk, eventually culminates in the robot’s convergence with the green target.

Integrated Perception Architecture

The oscillations of signal protein concentrations presented in the run and tumble decision architecture (FIG.23D) alternates the robot’s behavior by toggling a state machines encoding for wheel angular velocity, and thereby enacting robot translation or rotation. The robot’s behavior was observed by video and the motion path observed on the video is presented in Fig. 23A. Previously, this oscillation was produced by coupling a biased gene network with chemical pulses from the host processing hub. However, there are well-known gene networks that oscillate without the need for external chemical inducers³⁰. By programming the microbiome processing hub to contain one of these oscillating gene networks, an integrated perception architecture was developed (FIG. 24A), that used simulated biological components to short term memory, augmenting the memory hierarchy needed for image processing. The motion path found in FIG. 24A was taken from a video recording of the robot’s behavior.

The integrated perception architecture allowed for the microbiome processing hub to play a role in interpreting the robot’s environment, whilst maintaining control of the actuation presented in the run and tumble architecture. The control structure (FIG. 24B) was deceptively simple: elevated levels of signal protein one caused the robot to translate forward whereas elevated levels of signal protein two caused the robot to rotate. Additionally, an input chemical pulse of N-acyl homoserine lactone (AHL) was delivered to the microbiome processing hub at a rate proportional to the size of the target, as perceived by the robot’s onboard computer vision.  

The microbiome processing hub within the integrated perception architecture was driven by simulated cell population endowed with two regulatory gene networks. The first was an oscillating synthetic gene circuit that was synthetically developed close to ten years ago and has since become a canonical feature within the field of synthetic biology^{30, 31}. The second network was an orthogonal circuit containing negative feedback³². This gene network caused elevated levels of signal protein one only when levels of AHL within the cell increased, a condition. However, when the target is perceived as further away, causing AHL input to lessen, the circuit represses and concentration of signal protein one attenuates. These dynamics can be seen in (FIG.24D).

Within our experiment, AHL was pulsed to the microbiome processing hub in a concentration proportional to size of target image, thereby allowing the simulated cells to process whether or not the target was relatively closer. This had the affect of offloaded image memory from the robot’s embedded processing and onto the microbiome processing hub’s biochemical concentrations. This established a paradigm for using the microbiome processing hub not only as a computational actor, but also as an aspect of the memory hierarchy.

Exploring Environmental Variations

Having engineered three different robot control architectures linking the microbiome and host processing hubs, it was tested how cells may be used to make decisions for robots under varied environmental conditions.

First, by using the same control architecture and gene regulatory networks presented in FIG. 23B and 23C, a robot to alternate between seeking green or red targets was programmed (FIG. 25A). The motion path found in FIG. 25A was taken from a video recording of the robot’s behavior. This behavior commenced with sustained expression of signal protein one FIG. 25B that drove the robot to converge with the green target. At this moment, a pulse of a chemical associated with the green target (arabinose) was delivered to the microbiome processing hub. This pulse caused the gene regulatory network to flip its bistability, ceasing signal protein one expression and up-regulating signal protein two production. Correspondingly, the robot executed a biased random walk leading to an eventual convergence with the red target.

The inherent nature of the genetic toggle switch causes the robot to alternate between different targets. By changing the robot’s subroutine to a direct seek (FIG. 25C) rather than a biased random walk, we are able to program a resilient alternating behavior. The motion path found in FIG.25C was taken from a video recording of the robot’s behavior. The graph of the signal proteins (FIG. 25D) indicates a clear switch-like behavior following transient pulses of input chemicals simulated in the microbiome processing hub. It should be noted that the displayed alternating behavior captured in FIGS. 25A and 25C, in practice,  

continues beyond four targets; these figures and videos are meant to be characteristic snapshots the underlying robot dynamics.

Preferential Resource Selection

FIGS. 25A-25D presents a balanced, mutually repressible gene circuit that caused the robot to execute periods of sustained search. However, by biasing the gene network to one side, we observed timer-like effects in the signal protein concentrations, similar to those shown in FIGS. 23A-23D. The speed of this timing feature results from the strength of the bias in a gene network. This feature allows us to explore how decision hierarchies may be established by biological components within the microbiome processing hub.

Nature provides many examples where organisms will preferentially seek a certain nutrient based off of its relative fitness advantage. As such, we wanted to engineer our biomimetic robot system to seek targets with a preferential hierarchy governed by potentially mutable biological components within the microbiome processing hub.

An experiment was designed that placed the robot at the center of an arena containing three red targets and one green targete (FIG. 26A). The motion path found in FIG. 26A was taken from a video recording of the robots behavior. By modifying the control structure developed in FIGS. 22A-22D, we were able to create a decision architecture, similar to the master/slave architecture FIG. 26B), that mimicked preferential target pursuit. Crucially, by adding a feature that delivered chemical pulse to the simulated cells when the robot was in proximity to the red target, we introduced a vehicle for environmental feedback. This chemical pulse allowed to the host processing hub to “inform” the microbiome processing hub when it was close to a red target.

This robot’s behavior was driven by a simulated cell containing a biased toggle (FIG. 26C) similar to that used for the run and tumble architecture. In this simulation the toggle was biased producing elevated levels of signal protein one and low levels of signal protein two. Chemical (arabinose) pulses would cause a temporary attenuation of signal protein one while and elevating the concentration of signal protein two (FIG. 26D) until the bias of the toggle would restore elevated original stability.

This timing feature caused the robot to temporarily seek red targets, before returning to its initial objective of seeking the green target. Such a behavior is reminiscent of a predation habit in which one resource is preferentially sought. In the example chosen for FIGS. 26A-26D, we note two periods of secondary (red) target pursuit, one unsuccessful at time 60h and one successful around time 90h. These two time periods within this experiment show how preferential resource selection can result in wasted energy (time period 1) or a fitness advantage (time period 2) depending on the outcome of the secondary resource search.

 

Balancing these costs is computationally intensive for a simple organism. However, by varying the relative strength of the gene network’s bias, we can explore how a simple networks may evolve to optimize organism fitness in a target, and thereby resource, -rich landscape. By varying a genetic component that encodes for the amount of repression protein produced by the cells, we can alter the bias of the gene network. As this parameter increases, we observe (FIG.27) that the robot spends a larger percentage of time searching for the red targets. By mutating one evolvable genetic component, the genes within the microbiome processing hub causes the robot to behave in subtly different ways which may be energetically optimized if coupled with a fitness constraint. In such a way, FIG. 27 provides a framework for evolved task optimization³³.

Discussion

This Example demonstrates an engineered bioinspired robot system that allowed for hybrid decision architectures, incorporating inputs from a microbiome processing hub and a host robot processing hub. Using this system, it was explored how hybrid control structures could be employed to reproduce bioinspired behaviors for target acquisition. Additionally, methods for programming complex behaviors were developed, such as preferential resource selection, with simple regulatory gene.

Three different architectures were engineered for examining how dual processing hubs, inspired by host-microbiome systems, could interact. Not surprisingly, each of these decision architectures offered a distinct set of advantages. The master/slave paradigm had the fastest average time of convergence; the run and tumble took on average the fewest physical steps (7.4% less that master/slave and 6.5% less than the integrated perception); finally, the integrated perception was able to off-load aspects of the vision processing and memory from the host hub and into the microbiome hub. This sets up a collaborative memory architecture in which chemical signals act as information packets, bussed to the simulated microbiome.

The advantages are not trivial. Engineered cells have been shown to evolve in response to fitness constraints³⁴. Therefore, each of these control architectures offers a potential vehicle for performance optimization. A global sensitivity analysis^{35, 36} readily identified a mutable biochemical parameter, the lacI ribosome binding site strength, that dramatically affected the performance of the genetic toggle switch. When considering robot performance, varying the strength of this parameter affected the characteristics of a preferential resource selection behavior (FIG.27).

Importantly, our biomimetic robot and decision architectures serve as a proof of concept for any device with embedded electronics. The same dual processing living- nonliving interface we used to locomote our robot could just as easily control pumps or actuate valves. This raises the possibility of living populations evolving to optimize a broad  

range of tasks. The biomimetic gut-brain axis developed here can enable a range of scientific professions, from biologists seeking model systems for host-microbiome interactions to mechanical engineers exploring novel control structures and mechatronic optimization. References for Example 2

1. LeCun, Y., Bengio, Y., and Hinton, G. (2015) Deep learning, Nature 521, 436-444. 2. Howard, A. (2006) Multi-robot Simultaneous Localization and Mapping using Particle Filters, The International Journal of Robotics Research 25, 1243-1256.

3. Matthew, K., Karl, K., and Henry, W. (2012) Development of the Nano Hummingbird:

A Tailless Flapping Wing Micro Air Vehicle, In 50th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, American Institute of Aeronautics and Astronautics.

4. Shang, J. K., Combes, S. A., Finio, B. M., and Wood, R. J. (2009) Artificial insect wings of diverse morphology for flapping-wing micro air vehicles, Bioinspiration & Biomimetics 4, 036002.

5. Feifei, Q., Tingnan, Z., Wyatt, K., Paul, B. U., Robert, J. F., and Daniel, I. G. (2015) Principles of appendage design in robots and animals determining terradynamic performance on flowable ground, Bioinspiration & Biomimetics 10, 056014.

6. Marvi, H., Gong, C., Gravish, N., Astley, H., Travers, M., Hatton, R. L., Mendelson, J.

R., Choset, H., Hu, D. L., and Goldman, D. I. (2014) Sidewinding with minimal slip: Snake and robot ascent of sandy slopes, Science 346, 224-229.

7. Button, R. W., Kamp, J., Curtin, T. B., and Dryden, J. (2009) A survey of missions for unmanned undersea vehicles, DTIC Document.

8. Simaan, N., Taylor, R., and Flint, P. (2004) A dexterous system for laryngeal surgery, In Robotics and Automation, 2004. Proceedings. ICRA'04.2004 IEEE International Conference on, pp 351-357, IEEE.

9. Steingrube, S., Timme, M., Worgotter, F., and Manoonpong, P. (2010) Self-organized adaptation of a simple neural circuit enables complex robot behaviour, Nat Phys 6, 224-230.

10. Ayers, J., and Witting, J. (2007) Biomimetic approaches to the control of underwater walking machines, Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 365, 273-295.

11. Tan, Y., and Zheng, Z.-y. (2013) Research Advance in Swarm Robotics, Defence Technology 9, 18-39.

12. Markram, H.,et al.Reconstruction and Simulation of Neocortical Microcircuitry, Cell 163, 456-492.  

13. Silver, D., et al. (2016) Mastering the game of Go with deep neural networks and tree search, Nature 529, 484-489.

14. Montiel-Castro, A. J., González-Cervantes, R. M., Bravo-Ruiseco, G., and Pacheco- Lopez, G. (2013) The microbiota-gut-brain axis: neurobehavioral correlates, health and sociality, Frontiers in Integrative Neuroscience 7.

15. Bercik, P., Denou, E., Collins, J., Jackson, W., Lu, J., Jury, J., Deng, Y.,

Blennerhassett, P., Macri, J., McCoy, K. D., Verdu, E. F., and Collins, S. M. The Intestinal Microbiota Affect Central Levels of Brain-Derived Neurotropic Factor and Behavior in Mice, Gastroenterology 141, 599-609.e593.

16. Sharon, G., Segal, D., Ringo, J. M., Hefetz, A., Zilber-Rosenberg, I., and Rosenberg, E. (2010) Commensal bacteria play a role in mating preference of Drosophila melanogaster, Proceedings of the National Academy of Sciences 107, 20051-20056. 17. Heyde, K. C., and Ruder, W. C. (2015) Exploring Host-Microbiome Interactions using an in Silico Model of Biomimetic Robots and Engineered Living Cells, Scientific Reports 5, 11988.

18. Braitenberg, V. (1986) Vehicles: Experiments in Synthetic Psychology, Philosophical Review 95, 137-139.

19. Michaelis, L., and Menten, M. L. Die kinetik der invertinwirkung.

20. Garcia-Ojalvo, J., Elowitz, M. B., and Strogatz, S. H. (2004) Modeling a synthetic multicellular clock: Repressilators coupled by quorum sensing, Proceedings of the National Academy of Sciences of the United States of America 101, 10955-10960. 21. Elowitz, M. B., and Leibler, S. (2000) A synthetic oscillatory network of transcriptional regulators, Nature 403, 335-338.

22. Gardner, T. S., Cantor, C. R., and Collins, J. J. (2000) Construction of a genetic

toggle switch in Escherichia coli, Nature 403, 339-342.

23. Chandran, D., Copeland, W. B., Sleight, S. C., and Sauro, H. M. (2008) Mathematical modeling and synthetic biology, Drug Discovery Today: Disease Models 5, 299-309. 24. Ghosh, K. K., Burns, L. D., Cocker, E. D., Nimmerjahn, A., Ziv, Y., Gamal, A. E., and Schnitzer, M. J. (2011) Miniaturized integration of a fluorescence microscope, Nat Meth 8, 871-878.

25. Adler, J., and Tso, W.-W. (1974) "Decision"-Making in Bacteria: Chemotactic

Response of Escherichia coli to Conflicting Stimuli, Science 184, 1292-1294.

26. Darnton, N. C., Turner, L., Rojevsky, S., and Berg, H. C. (2007) On Torque and

Tumbling in Swimming Escherichia coli, Journal of Bacteriology 189, 1756-1764. 27. Fujinami, S., Terahara, N., Krulwich, T. A., and Ito, M. (2009) Motility and chemotaxis in alkaliphilic Bacillus species, Future Microbiology 4, 1137-1149.

 

28. Khalil, A. S., and Collins, J. J. (2010) Synthetic biology: applications come of age, Nature Reviews Genetics 11, 367-379.

29. Ellis, T., Wang, X., and Collins, J. J. (2009) Diversity-based, model-guided

construction of synthetic gene networks with predicted functions, Nat Biotech 27, 465-471.

30. Stricker, J., Cookson, S., Bennett, M. R., Mather, W. H., Tsimring, L. S., and Hasty, J. (2008) A fast, robust and tunable synthetic gene oscillator, Nature 456, 516-519. 31. Hasty, J., Dolnik, M., Rottschäfer, V., and Collins, J. J. (2002) Synthetic Gene

Network for Entraining and Amplifying Cellular Oscillations, Physical Review Letters 88, 148101.

32. Voliotis, M., and Bowsher, C. G. (2012) The magnitude and colour of noise in genetic negative feedback systems, Nucleic Acids Research 40, 7084-7095.

33. de Smit, M. H., and van Duin, J. (1990) Secondary structure of the ribosome binding site determines translational efficiency: a quantitative analysis, Proceedings of the National Academy of Sciences 87, 7668-7672.

34. Buller, A. R., Brinkmann-Chen, S., Romney, D. K., Herger, M., Murciano-Calles, J., and Arnold, F. H. (2015) Directed evolution of the tryptophan synthase β-subunit for stand-alone function recapitulates allosteric activation, Proceedings of the National Academy of Sciences 112, 14599-14604.

35. Feng, X.-j., Hooshangi, S., Chen, D., Li, G., Weiss, R., and Rabitz, H. (2004)

Optimizing Genetic Circuits by Global Sensitivity Analysis, Biophysical Journal 87, 2195-2202.

36. Zhang, X. Y., Trame, M. N., Lesko, L. J., and Schmidt, S. (2015) Sobol Sensitivity Analysis: A Tool to Guide the Development and Evaluation of Systems

Pharmacology Models, CPT: Pharmacometrics & Systems Pharmacology 4, 69-79.

 

Claims

We claim: 1. A system comprising:

a microbiome, wherein the microbiome comprises a genetically engineered bacterium that comprises a synthetic gene circuit responsive to an inducer and wherein the synthetic gene circuit is configured to generate an optically active protein;

a microscope, wherein the microscope is optically coupled to the microbiome and wherein the microscope is configured to receive an optical output from the optically active protein and produce an output voltage correlating to the optical output from the microbiome;

processing circuitry comprising a processor and a memory, wherein the processing circuitry is coupled to the microscope; and

an application comprising machine readable instructions stored in the memory that, when executed by the processor, cause the processing circuitry to at least:

receive the output voltage from the microscope;

determine a characteristic of the output voltage; and

provide a command to a robotic device to execute a function based at least in part upon the output voltage.

2. A system comprising:

a microbiome, wherein the microbiome comprises a genetically engineered bacterium comprising a synthetic gene circuit responsive to an inducer;

a sensor, wherein the sensor is coupled to the microbiome and wherein the sensor is configured to produce an output signal corresponding to the biologic activity of the microbiome detected by the sensor;

processing circuitry comprising a processor and a memory, wherein the processing circuitry is coupled to the sensor; and

an application comprising machine readable instructions stored in the memory that, when executed by the processor, cause the processing circuitry to at least:

receive the output signal from the sensor; and

provide a command to a robotic device to execute a function based at least in part upon the output signal.

.

 

3. The system of claim 2, wherein the sensor is optically, biologically, chemically, fluidically or electrically coupled to the microbiome.

4. The system of claim 2, wherein the processing circuitry is electrically, optically, or wirelessly coupled to the sensor.

5. The system of claim 2, further comprising the robotic device, wherein the robotic device is electrically, optically, or wirelessly coupled to the processing circuitry.

6. The system of claim 2, further comprising a chemostat, wherein the microbiome is contained in the chemostat.

7. The system of claim 6, further comprising the robotic device, wherein the robotic device is coupled to the chemostat.

8. The system of claim 7, wherein the robotic device is electrically, optically, or wirelessly coupled to the chemostat.

9. The system of claim 6, wherein the chemostat is electrically, fluidically, or optically, coupled to the sensor.

10. The system of claim 6, wherein the chemostat comprises a microfluidic channel, wherein the microbiome is contained in the microfluidic channel.

11. The system of claim 6, wherein the chemostat is a microchemostat.

12. The system of claim 2, further comprising a microscope that is optically coupled to the microbiome and is electrically, optically, or wirelessly coupled to the sensor.

13. The system of claim 12, wherein the synthetic gene circuit is configured to generate an optically active protein and the biologic activity of the microbiome sensed by the sensor is a wavelength of light produced by the optically active protein.

14. The system of claim 2, wherein the application, when executed by the processor, additionally causes the processing circuitry to at least determine a characteristic of the output signal.

 

15. The system of claim 2, wherein the output signal is a voltage, optical signal, chemical signal, biologic signal, audio signal, or electromagnetic signal.

16. The system of claim 2, wherein the characteristic is the average output signal over a period of time.

17. The system of claim 2, wherein the function is to move the robotic device in a direction over a distance.

18. The system of claim 2, wherein the inducer is an environmental inducer.

19. A method of controlling a robotic device, the method comprising;

growing a microbiome in a chemostat, wherein the microbiome comprises a genetically engineered bacterium comprising a synthetic gene circuit responsive to an environmental inducer;

sensing a biological activity of the microbiome;

generating a signal in response to the sensed biological activity; and

transmitting a command to a robotic device to execute a function based at least in part upon to the output voltage.

20. The method of claim 19, wherein the microbiome is grown such that substantially all of the bacteria of the microbiome are in the exponential growth phase.

21. The method of claim 19, wherein the synthetic gene circuit is configured to generate an optically active protein.

22. The method of claim 21, wherein the biological activity is light generated by the optically active protein.

23. The method of claim 18, wherein the signal is voltage, optical signal, chemical signal, biologic signal, or electromagnetic signal.

24. The method of claim 18, wherein the chemostat is a microchemostat.