Phase 1 Results
The first phase of our project was to develop a library of digital parts. Our journey began with finding the guide RNA sequences we would use in our work. We generated over one thousand 20 base guide RNA sequences in silico using a random sequence generator.
It was critical these guide RNA sequences be orthogonal to the human genome because we wanted to test our system in HEK293FT cells and prevent off target activation in the host genome. To test for orthogonality, we entered the sequences into the CRISPR Optimized Design tool developed by the Feng Zhang lab. We selected the top 18 sequences, which had an orthogonality score of 98% or higher, to synthesize and use in our research.
We synthesized these guide RNA sequences through IDT and cloned each one into guide RNA expressing vectors and guide RNA operator pairs. To act as controls for our experiments, we also cloned the guide RNA sequences Tre and UAS into guide RNA expressing vectors and guide RNA operator pairs. These 20 initial operators expressed an iRFP gene.
We transiently transfected and ran through a flow cytometer our paired gRNA expression vectors and gRNA operator reporters and dCas9-VPR to test their expression behaviors. We wanted to see high activated states coupled with low basal expression, a true digital system. The screen of this can be seen below.
Based on this screen, we decided to proceed with guide 1, guide 3, guide 8, and guide 13 for all future experiments. Next, we needed to prove that our parts could drive a diverse library of genes of interest. To complete this, we replaced the iRFP in our guide RNA operator reporters with a GFP, a BFP, and an mRuby protein. The results from these test can be seen in the matrix below.
In all cases, there was low basal expression and strong activation, proving that our system not only behaves digitally, but that it could also do so over multiple genes.
Finally, our system needed to be mutually orthogonal to prevent significant undesired crosstalk. If there was off target gene activation between operators, then we could not transfect multiple operator-expression pairs and obtain predictable results. We performed an orthogonality transfection with the GFP driven operators from our library. The experiment was designed such that every guide RNA expressing vector would be paired with each guide RNA operator reporter vector. We predicted significant GFP fluorescence when the guide RNA expressing vector has the same guide RNA sequence as the guide RNA operator reporter vector, and no significant expression elsewhere. The results of the experiment can be seen in the matrix below.
The diagonal down the matrix demonstrates that there is mutual orthogonality between the guides in our system. With this matrix, we proved that we developed a digital parts library.
Our final test was to compare the relative strength of our parts to a CMV, a strong mammalian promoter, to determine the relative strength of our system. This would also allow for a better interpretation of where our parts fit into the grand scheme of synthetic biology. The graph below shows the results from this experiment.
While our parts expressed well, they did not have a higher expression than the CMV. This was not particularly surprising. But it did raise the question, would we ever be able to show higher activity with a minimal CMV in our system than with a full CMV?
Our next step was to expand to a functioning analog library...
Phase 2 Results
The next step in developing our library was to expand into the analog domain. We are more specifically referring to an array of quantized levels of expression that on a macroscopic scale appear analog, similar to a Riemann Sum of an analog signal.
To create this “quantized-analog” pattern, we had two driving hypotheses: multimerize the operators to increase expression and mutate the operators to decrease expression.
The conceptualization behind multimerizing, or adding multiple binding sites for the dCas9, harkens to previous work done with TAL effectors and Zinc Finger motifs. By adding additional binding sites, we expect to recruit more dCas-VPR transactivators to the operator vector, increasing expression and producing a synergistic effect. Additionally, we varied the space between binding sites. The motivation behind this was that we wanted space for the dCas9-VPR to bind effectively. The results for these experiments for each of the four guides in our library can be seen in the graphs below.
In general, increasing the number of binding sites also increased the expression of the operator reporter. Additionally, with only minor variations from this trend, increasing the space between binding sites also increased expression. This set of experiments proves that multimerization is an effective technique to increase gene activity.
We also compared the relative strength a CMV from the Registry to one of our multimerized operators, consisting of three binding sites for guide RNA 13. CMVs are known among mammalian synthetic biologists as being a strong constitutive promoter. The data from this comparison can be seen below.
As the graph demonstrates, our triple multimerized operator containing a minimal CMV had greater expression than the full CMV. This proves that our system produces additional varied levels of gene activation as well as have some temporal inducibility.
After we completed cloning and assaying our multimerized operator reporters we sequentially mutated every base in our twenty base guide RNA sequences, beginning at the five prime end. Every adenosine was mutated to a cytosine and vice versa and every guanine was mutated to a thymine and vice versa. This produced a complete shift from purines to pyrimidines of the opposite DNA base pair. In addition to these single base mutations, we also clustered 2 and 3 point mutations around base 1 and base 11 in the guide. The results of this screen can be seen below.
Though the data did not produce the decreasing step function we had expected, there were some results of extreme interest. The mutations on bases 4 through base 7 produced a distinguishable downward trend. Moreover, mutations at base 10 and base 11 produced nearly a fivefold decrease in expression when compared to the non-mutated operator. From our multi point mutations, the data did not represent any significant or controllable changes in expression. Therefore, we have decided to not move forward with multi-point mutations.
With this experiment we have proven that we have developed an analog parts library. We have a system than can range from five times below normal expression to five times above normal expression under a minimal promoter, out competing a strong full promoter. Now, what can we accomplish with our library of both digital and analog parts?
An answer lies in Genetic Logic Circuits...
Phase 3 Results
As we move to combat more and more complicated problems in synthetic biology, we must create more and more advanced tools. As synthetic biologists, one route we can take is to develop genetic logic circuits to produce the computational tools needed to complete our research.
There are several possible variants of genetic logic circuits. We chose to use combinatorial logic circuits because it is only the absence or presence of inputs, not the order, that determines the output.
By definition, these inputs shift the logical state of our circuits. To build our circuits, the guide RNA target sequences were placed between heterospecific recombination sites unique to Cre and Flp tyrosine recombinases. When a recombinase is introduced, it excises the region between the heterospecific sites, changing which guide RNA is expressed through the constitutive promoter that drives the circuit. It must be noted that this process is unidirectional in time, meaning that once a state change has been made, it cannot be undone.
We performed several experiments of increasing complexity to demonstrate that our library could be used a wide variety of circuits, with digital and analog outputs. All of these experiments, which will be shown below, used the same gene circuit. That circuit can be seen in the figure below:
The circuit constitutively expresses guide RNA 3 in the first state, then expressed guide RNA 1 in the second state, followed by guide RNA 8 in the third state, and finally guide RNA 13 in the final state. This image will appear in conjunction with every experiment below as a reminder.
Experiment 1: Activation
One of the most fundamental behaviors of genetic logic is to be able to activate a gene. We set up an AND gate experiment to test this behavior. AND gates are defined as only producing an output in the presence of both signals. To do this we transfected our logic circuit as well as a guide RNA operator reporter from the Gemini library that had a g13 target site and a BFP reporter. The expectation was that no significant fluorescence should be detected in state 1, 2, or 3. However, in state 4 we should see strong expression of BFP. As the graph below demonstrates, we succeeded in proving our system could make an AND gate.
Experiment 2: Repression
A second important fundamental behavior of genetic logic circuits is to repress the expression of a gene. To demonstrate this, we developed a NOR gate experiment. A NOR gate is defined as only producing expression when there is no signal present. To do this we transfected our logic circuits as well as a guide RNA operator from the Gemini library that had a g3 target site and a GFP reporter. The expectation was that there would only be significant expression of GFP in the state 1, and no significant expression in the other three states. As the graph below demonstrates, we succeeded in proving our system could make a NOR gate.
In addition to the flow cytometry results, we sent our circuit our to Worcester Polytechnic Institute’s iGEM team to validate our results using fluorescent microscopy. The images from the microscope can be seen below.
Examining the images, you can see that in the first state there is significant expression of GFP in state 1 and there is no significant expression of GFP in the other three states. This further proves the above claim.
Experiment 3: Complexation of Basic Gates
Once we were able to proof our library’s ability to perform fundamental circuits behaviors, we moved to develop more complex behaviors. In this case, we combined our AND and NOR to develop our “Complexation Circuit”. The expected behavior of our circuit was that in absence of signals or in the presence of both signals there would be expression. From Boolean perspective that means 0/0 input or 1/1 input. To perform this experiment, we transfected our circuit as well as the operators from our AND gate and our NOR gate experiment. In state 1 we expected to see an expression of GFP and in state four we expected to see BFP be expressed. There should be no significant expression of either protein in states 2 and 3. As the results below show, our “Complexation Circuit” functions as expected, proving our circuits capacity to perform more complex functions than fundamental computation.
Experiment 4: Line Decoders
The final test for a digital output circuit we wanted to perform was to develop a line decoder. Line decoders represent the quintessential circuit behavior. They are defined by each logical state having its own output. We performed this experiment four distinct times, having each state produce each of our four fluorescent proteins (GFP, BFP, mRuby, and iRFP). The results of these four trials can be seen below with the operators used in the experiment stated:
g3- BFP g1- GFP g8-iRFP g13- mRuby
g3- GFP g1- iRFP g8- mRuby g13- BFP
g3- mRuby g1-BFP g8- GFP g13-iRFP
g3- iRFP g1- mRuby g8- BFP g13- GFP
In each case there is only significant expression of a given fluorescent in its appropriate state. All in all, this proofs our library can handle a complex system of four inputs and four outputs.
Experiment 5: Analog Logic
Once we were able to proof our system could handle four distinct outputs we moved to test an analog output circuit. As a reminder, an analog output circuit would have the same gene of interest expressed at different levels given the specific logical state. Specially, we set up the experiment such that state 1 would pair with the triple multimerized plasmid with a g3 target site. State 2 would pair with a single binding site operator for a g1 target site. State 3 would pair with the double multimerized plasmid with a g8 target site. And finally, state 4 would pair with a mutant g3 target site. All of these operators would drive the expression of GFP and would be transfected with the genetic logic circuit. The expected results of this experiment were as follows: State 1 would expresses the highest, followed by State 3, followed by State 2, and finally State 4. The results of the experiment can be seen in the graphic below.
While there are some discrepancies between the expected results and the actual results, (specifically the expression level of State 3), we can say that this experiment effectively proves that the Gemini library of parts can be used to produce analog response from a combinatorial circuit.