Team:Thessaloniki/Design
Overview & Inspiration
"What I cannot create I do not understand." -Richard Feynman
In late May we as a team got besotted by the workings of molecular programming techniques and the ways that nucleic acids can be programmed to perform logic operations. We sought out to uncover the various arrangements that experts in the field have formulated over the years to achieve such operations.
Bewitched by the many applications of DNA strand displacement, its elegance, and potentialities, we were determined to find a novel way to make use of the aptitudes of molecular programming to intensify advances in basic fields of biological research.
The terrain of DNA-protein interactions was a perfect match. The soaring need to elucidate such interactions in the fields of diagnostics, therapeutics as well as cell and systems biology, rendered the need for their quantitative study of utmost importance.
In that accord, we sought out to entangle DNA-protein interactions into the molecular computation performed through nucleic acid molecules, in the form of a novel formal Chemical Reaction Network (CRN) that takes both DNA molecules and DNA-binding proteins as inputs while producing measurable output.
We experimentally demonstrate a design strategy for building DNA-only chemical circuits capable of being programmed to execute analog temporal dynamics. The technology is designed around a signaling protocol based on short single-stranded DNA sequences. The control system we design receives inputs in the form of DNA sequences and produces outputs in the form of other sequences.
Our DNA components are, in principle, capable of realizing the entire diversity of dynamic behaviors of chemical kinetics as mathematically captured by a chemical reaction network (CRN).
Our experiments corroborate that we can realize complex behaviors previously out of reach of synthetic molecular systems.
Chemical Reaction Network
Like every computer’s processing unit, DNA logic gates interact with each other based on a certain architecture. In molecular programming, this architecture can be expressed in the language of Chemical Reaction Networks. A CRN, in our case, defines a set of nucleotide species and the set of reactions those species can do, and the CRN model allows us to deduce the global behavior of the system based on that local definition.
Although CRNs started out as a method of understanding experimental observations produced by elementary chemical reactions, they are used in forming a general framework for modeling systems with multiple interacting components, such as gene regulatory networks, animal populations, and sensor networks. CRNs can embody a wide range of digital and analog behaviors, including temporal pattern generation, multistability, and memory, Boolean logic, signal processing, control systems or distributed algorithms. Moreover, viewed as a programming language, CRNs provide a natural and intuitive format for the reasoning and description of molecular interactions, without making the underlying physical details explicit.
We set to design and program our own CRN in order to make it applicable for DNA-protein interaction quantification as well as providing elements of orthogonality and robustness in a microenvironment of protein-DNA interactions.
Our initial approach was directed towards an amplification scheme that produced amplifiable output readout in time and input dependent manner.
We came across two possible choices for this amplification to be facilitated. An autocatalytic and a catalytic CRN.
In the autocatalytic amplification scheme, the input to the CRN would be set in the form of the missense of protein binding. This output would then be 'translated' into an autocatalytic single-stranded DNA that would promote the production of excess amounts of output in a small amount of time, predictably. [1][2]
However, through our model simulations and estimations (link) we came to the conclusion that such a system would not be easily adaptable and a much thorough design must have been implemented. Additionally, DNA processing gates must have been exclusively synthesized, thus making this implementation troublesome while lacking modularity.
Experiencing these hindrances, we came to the conclusion that a much modular and accessible way of receiving DNA logic gates must be pursued. Inspired by the applicability and modularity of synthetic biology techniques, as well as the accessibility and robustness of DNA vectors and the straightforwardness of such DNA extraction and purification, we focused on devising a method of establishing a formal CRN based on bacterial plasmid DNA. We designed DNA computing gates that could be derived from circular double-stranded DNA after simple processing steps with standard molecular biology techniques.
Our design involves linear DNA Strand Displacement gates that could be derived out of plasmid DNA through a standard restriction reaction with a blunt end cutter restriction endonuclease (PvuII) and the site-specific nicking of the upper DNA strand by nicking endonucleases (Nt.BstNBI/Nb.BsrDI).
That way we implement a method of receiving cost-effectively DNA computing gates. Plasmid DNA is considered high quality as the intracellular mechanisms minimize sequence errors, mismatches, and mutations making our DNA logic gates error-proof while their supply could be endless as bacteria could be stored in glycerol stocks for over 10 years.
For our amplification scheme, we settled on a catalytic CRN that can be expressed through the familiar language of chemistry that can be used to write programs for our DNA architecture. The ‘instruction’ A+B→ C+B that we have decided upon means that the signals A and B are transformed into signal C, where A, B, and C are DNA strands we design. The reaction is not elementary; rather, it is systematically ‘compiled’ into a sequence of DNA strand displacement reactions.
We designed possible implementations that could have the intended function and programmed them into molecular programming language to evaluate through simulations of their behavior. Our final design implementation resulted in a collection of three linear logic gates and a reporter complex that all receive the upstream gate’s output as input.
We designed our Input Gate to be an upstream gate that feeds input TBB to the Join Gate. Join Gates then act as the connection between the Input and the Fork Gates. Subsequently, the Fork Gate produces both the output of the circuit and another TBB input, facilitating the intended amplification. This design ensures that even very small amounts of Input Gate will eventually consume all available Join and Fork Gates, producing a predictable measurable output that is deviating in regards to the time needed for the intended amplification to take place.
Linear Amplification Algorithm Flowchart
Our CRN’s architectu re is initiated in the upstream Input Gate. The reaction that becomes the trigger for the circuit is defined as the moment auxiliary strand DTG displaces strand TDD of the Input Gate.
That way, toehold TD becomes exposed and ready to be bound by the next auxiliary strand BTD that will eventually free the input strand TBB from the Input Gate.
Finally, auxiliary strand NTB will conclude strand displacement reactions on the Input Gate, rendering the complex fully double-stranded, without any exposed toeholds that could potentially make these reactions bidirectional.
At the Join Gate, auxiliary strand TAA initiates the strand displacement reactions displacing the nicked strand ATB, thus freeing toehold TBB.
The above step is necessary for the gate to be prepared to receive the input TBB from the upstream Input Gate. The presence of this strand initiates the displacement of the nicked strand BTR and frees toehold TR, allowing for the auxiliary strand TRR to conclude the strand displacement reactions for the Join Gate. Moreover, the nicked strand BTR will be used in downstream reactions as well.
After strand TRR displaces the nicked strand RTQ, it acts as an initiator for the strand displacement cascade of the Fork Gate. This process produces an additional auxiliary strand TRR which helps augment the amplification.
Subsequently, the upstream nicked strand BTR that was freed up earlier, acts as an auxiliary strand that continues with the interrelation between Join and Fork Gates. From that process, Input strand TBB, which was designed to be part of the gate, is freed exposing its designated toehold. This is the pivotal step of our CRN which facilitates the intended amplification. Through that step, the circuit’s input is refreshed, allowing even trace amounts of input (and consequently Input Gates) to consume all Join and Fork Gates
For the readout of the circuit’s behavior, the next strand displacement step produces the circuit’s output TCC with the help of auxiliary strand CTB. This input will be used downstream to make the circuit’s behavior measurable. Finally, auxiliary strand ITC concludes the strand displacement reactions, rendering the fork gate fully double-stranded, without exposed toeholds, forcing the CRNs trajectory in only one direction.
The readout of the circuit’s output can be facilitated by a displacement reaction of the TCC output strand on a reporter complex. This complex is designed to incorporate a quencher molecule on the 5’ end of its end that quenches the fluorescence of a fluorophore that is attached to the 3’ end of its corresponding bottom strand. When output TCC is released, however, the displacement of the quencher-carrying top strand frees up the fluorophore strand making a measurable fluorescent signal that follows the kinetic behavior of our circuit.
This year, our team designed a novel readout method for the detection of the circuit’s output never before used to measure strand displacement reactions. We deployed a Field Effect Transistor (FET) that measures voltage and current amplitude data derived from a gold plated electrode. The electrode is coated with DNA probes that act as our reporter complex. The difference in current amplitude is detectable due to the presence of a redox tag (in our case methylene blue) that is attached to one of the probed DNA strands, facilitating the exchange of electrodes with the gold surface of the electrode. This creates a distinct increase in current amplitude at a certain voltage spectrum. However, when the strand displacement is concluded, the release of the DNA strand carrying the redox tag terminates that exchange of electrodes, dimming the current amplitude.
Plasmid Derived Logic Gates
We designed the gate sequences so that they incorporate standard recognition sites for the PvuII restriction endonuclease and the Nt.BstNBI and Nb.BsrDI nicking enzymes respectively. Specifically, Input and Fork Gates were designed to be nicked with the Nt.BstNBI nicking enzyme, Join Gates incorporated the Nb.BsrDI nicking enzyme recognition site and all three gates had the site of PvuII incorporated at each end.
With that design, it becomes possible for the gates to be derived from plasmid DNA and through the appropriate enzymatic processing be prepared for DNA Strand Displacement Reactions.
The gates are designed to be orthogonal and fully modular. Their sequences come from our Genetic Algorithm model which is programmed to incorporate the appropriate enzyme recognition sites.
Plasmid derived gates, as mentioned above, are a great source of high-quality DNA to be used in molecular computation, reduce the cost of producing DNA Strand Displacement gates and minimize the error-prone synthesis and assembly method. Through that, possible leak pathways that appear because of those errors can be reduced as shown in [3] and the circuit leakage minimized.
In order for our CRN to receive protein-DNA binding as input, we designed our Input Gate to incorporate our designated Transcription Factor’s binding site sequence into the TDD domain.
As our input gate produces and feeds input TBB into the downstream Join Gate, the binding of a TF to its binding site will retard its ability to participate in strand displacement cascades. Thus the amount of the available input for downstream reactions will proportionately decrease relative to the binding affinity of the TF.
In order for the efficient incorporation of that non-DNA input in our DNA-only CRN, the TFs binding capability to specific DNA sequences must have been evaluated in order for the CRN to function as intended. Thus, we added the TFs binding site into the TDD domain of the Input Gate. However, it is also important for the intended purpose that no other possible TF binding sites are located into the other logic gates, so we tweaked our Genetic Algorithm model is designed to account for the fact.
Lastly, to test the specificity of our toolkit and it’s ability to distinguish between the binding affinity of a TF to varying binding sites, we inserted SNPs to the binding sequences through site-directed mutagenesis propagated by specific PCR primers. In principle, these SNPs would alter the Binding Affinity of the TF in regards to a single base bare change in its binding site.
[1] Zhang, D. Y., Turberfield, A. J., Yurke, B., & Winfree, E. (2007). Engineering entropy-driven reactions and networks catalyzed by DNA. Science, 318(5853), 1121-1125.
[2] Wang, B., Thachuk, C., Ellington, A. D., Winfree, E., & Soloveichik, D. (2018). Effective design principles for leakless strand displacement systems. Proceedings of the National Academy of Sciences, 115(52), E12182-E12191.
[3] Chen, Y. J., Dalchau, N., Srinivas, N., Phillips, A., Cardelli, L., Soloveichik, D., & Seelig, G. (2013). Programmable chemical controllers made from DNA. Nature nanotechnology, 8(10), 755.
[4] Li, Chao, et al. "In Vitro Analysis of DNA–Protein Interactions in Gene Transcription Using DNAzyme-Based Electrochemical Assay." Analytical chemistry 89.9 (2017): 5003-5007.
[5]
Introduction
Most assays directed towards the quantification of a DNA circuit’s output require the use of sophisticated instrumentation able to detect the overtime intensity of fluorescence proteins that indicate the activation of the circuit. However, as DNA has emerged as a powerful tool able to assemble complex molecular systems, it is attracting the interest of a continuously expanding audience in the scientific community, declaring a need for more accessible methods of DNA computing utilization. Therefore, new methods relying on electronic devices have been proposed, exploiting their versatility and efficiency in providing accurate results in a quick and reliable manner[1-4]. To that extent, our team developed an electrochemical assay of output measurement that relies on a gold sensing pad and is able to distinguish between the active and inactive states of our molecular circuit.
The design of this assay was based on newly emerged methods of nucleic acid hybridization sensing [3,4]. The construction of these sensing devices can be achieved using commercially available parts and without the need for any special equipment, declaring them an appealing approach to output detection.
We created a custom sensing pad and used it for recreating the methods proposed in [3] and [4]. The first experiment uses the sensing pad to measure capacitance variance [3], while the second uses it to measure conductance variance [4].
As proposed, PNA strands with a specific thiol acid tail can be immobilized on a golden electrode to act as probes, based on the sulfur-gold interaction. As shown (figure 1), using this electrode to form a capacitor measurement of the concentration of a specific DNA strand (supplementary to the probe) is possible, as the capacitance of the capacitor varies in dependence with the number of free probes. The capacitance value is converted to an analog voltage, using a depletion MOSFET, and the concentration of the specific strand can be measured using a voltmeter.
The second experiment in a similar way, measures the conductivity of the solution using a REDOX tag that affects the conductance of the electrode (figure 2). The electrode is coated with DNA probes that act as our reporter complex. The difference in current amplitude is detectable due to the presence of a redox tag (in our case methylene blue) that is attached to one of the probed DNA strands, facilitating the exchange of electrodes with the gold surface of the electrode. This creates a distinct increase in current amplitude at a certain voltage level. However, when the strand displacement is concluded, the release of the DNA strand carrying the redox tag terminates that exchange of electrodes, dimming the current amplitude. We used the internal Analog to Digital Converter (ADC) and Digital to Analog Converter (DAC) of a microcontroller to observe this fenomena.
Sensing pads construction
Using NI Ultiboard Student edition 14, a CAD software for designing Printed Circuit Boards, we designed a custom set of sensing pads (figure 3) and ordered it for manufacture at JLC PCB. Our electronic board was of simple design with gold finish (ENIG-RoHS 200um) and the total cost for 10 boards was less than 50$ (figure4). The electrodes of the capacitors are both made of gold, instead of AgCl for the anode as proposed [3,4]. This shortens the life expectancy of the anode electrode but also reduces the total cost of the sensor.
The Ultiboard project file can be found in
Also, the Gerber files for ordering the PCB can be found in
Capacitance measurement
For measuring the capacitance we constructed the proposed in [3] circuit (Figures 5, 6). Our measurement relies on the internal gate capacitance of the depletion MOSFET, creating a capacitor divider that is sensitive in any capacitance change. The MOSFET connected as a voltage follower is amplifying the gate’s voltage in order to be measured without affecting the balancing point of the capacitors.
In an attempt to further reduce the total cost of the measuring device we used an Arduino due board (Figure 7) to read the circuit’s output and convert it to capacitance. The circuit we used is depicted in its complete form in Figure 8.
The following parts were used:
- Arduino DUE (40$)
- IXTY08N100D2 (depletion MOSFET) (2$)
- Custom made gold plated pads (45$ 10boards x 100wells)
- R=1,8KΩ (0.1$)
- Potentiometer 10KΩ (0.5$)
- 9V Battery (4$)
Total cost 90$
The capacitance value is measured by the internal Analog to Digital Converter (ADC) of the Arduino board and then sent to a personal computer through serial connection (USB). Figure 9 shows the flow of the software running on the Arduino board. The code can be found in
For reading the measurements captured by the Arduino we used the following scripts found in our git repository. In the raw measurements obtained during the calibration of the system, as well as a python script that plots them can be found. The script is using pandas and matplotlib libraries. Also, for live viewing and saving the data we made another python script (using pandas, matplotlib, numpy, pyserial and drawnow libraries) that can be found in . The script obtains the Arduino data in real-time, converts the units (into volt and second) and plots a graph. Closing the graph window allows the user to save all obtained data into a CSV file.
Conductance measurement
For measuring the conductance, we used the internal Analog to Digital Converter (ADC) and Digital to Analog Converter (DAC) of the Arduino due board. The DAC is applying various voltages on the sensing pad, and the ADC measures the current flow, by measuring the voltage drop on a series-connected resistor. The circuit is shown in figure 10.
The following parts were used:
- Arduino DUE (40$)
- Custom made gold plated board (45$ 10boards x 100wells)
- R=1KΩ (0.1$)
Total cost 85$
The current value is measured by the ADC input of the Arduino and then sent to a personal computer through serial connection (USB). Figure 11 shows the flow of the software running on the Arduino board. The code can be found at .
For reading the measurements we used the serial monitor provided by the Arduino IDE software and plotted the curves using Excel. A workbook containing measurements during the calibration of the system can be found in .
References
[1] Arun Kumar Manoharan et al. Simplified detection of the hybridized DNA using a graphene field-effect transistor, Science and Technology of Advanced Materials (2017).
[2] Veigas B, Fortunato E, Baptista PV. Field-effect sensors for Nucleic Acid detection: recent advances and future perspectives. Sensors (Basel). 2015;15(5):10380–10398. Published 2015 May 4.
[3] Kaisti Matti, Kerko Anssi, et al. Real-time wash-free detection of unlabeled PNA-DNA hybridization using discrete FET sensor. 2017/11/16
[4] LI, Chao, et al. In Vitro Analysis of DNA–Protein Interactions in Gene Transcription Using DNAzyme-Based Electrochemical Assay. Analytical chemistry, 2017, 89.9: 5003-5007.
Transcription Factors:
Transcription factors, can modulate the rate of gene transcription through their ability to bind to DNA-regulatory sequences and result in over - or underexpression of protein synthesis, and subsequent altered cellular function. [1] The significant breakthroughs about their protein biochemistry and interactions with DNA at the structural level, combined with increasing needs for new targeted-approaches particularly in cancer, have openned the way for realizing the significance of characterizing and targeting transcription factors.
The potent targeting of transcription factors requires a better knowledge of their specific DNA binding sequences and high throughput method of quantifying their concentration in a sample.
Comprehending the importance of providing the scientific society with a toolkit that can fill the current gap in predicting transcription factors’ binding sites via in silico models, we have utilized the DNA Strand Displacement (DSD) mechanism to quantify DNA - protein interactions. Our circuit, as presented in the Chemical Reaction Network, has the ability to catalyze the anticipated interactions regarding both DNA molecules and DNA-binding proteins as inputs, while producing measurable output. The output produced comes in the form of fluorescence and differs in each of the aforementioned cases. In the case of those macromolecular compounds being attached on the proper DNA binding sequence, the fluorescent output is significantly lower than in the case of having just DNA molecules as input.
It becomes clear that genetic modifications of transcription factors, such as certain SNPs can have profound effects on the DNA-binding capacity of a transcription factor [2]. Therefore aiming at further quantifying the transcription factors’ binding, we tested their binding affinity at both consensus and modified with various SNPs sequences and measured the fluorescence outcome. The results from our measurements are comparably presented in the Results tab.
Regarding the selection of the transcription factors that were tested as a proof of concept for our project, we concluded in verifying our toolkit’s function with the p65 subunit of the NF-κB as well as with the ELK1 transcription factor. Both these factors play a pivotal role in the progression of metastatic melanoma and therefore their study and further characterization was estimated to be beneficial for the scientific community. Concerning NF-κB, it is considered as one of the most well characterized transcription factors, regulating many signal pathways. Therefore, its testing was basically aiming at validating our toolkit’s function. To the contrary, ELK1 was chosen as a possibility of further characterizing a crucial transcription factor for metastatic cancer.
[1] Ian M. Adcock, Gaetano Caramori, in Asthma and COPD (Second Edition), 2009
[2] Perkins ND . Post-translational modifications regulating the activity and function of the nuclear factor kappaB pathway . Oncogene 25 : 6717– 30 , 2006 .