Wet Lab Cycle

Creation and Characterization of a dsRNA Recognition Module

Initial Plan
We decided to use a truncated form of full-length RIG-I, in which the two CARD domains in N-terminal were removed, as a dsRNA-binding domain. Our goal is to create genetically engineered cells that express COCCO. However, during the development process, it was necessary to produce recombinant proteins to validate the function of individual parts. Specifically, to verify whether this domain binds to dsRNA as expected, we conducted in vitro assays in which purified recombinant protein was incubated with synthetic dsRNA. The protein in question is a long polypeptide of 85.6 kDa, and making it amenable to recombinant purification required multiple iterations of an engineering cycle.
Cycle1

Design

We initially planned to simply clone the corresponding region from the RIG-I cDNA and insert it into an expression vector, using Escherichia coli as the expression host.

Build

We designed a plasmid by inserting the sequence encoding an N-terminally His-tagged RIG-I-CARDapaf1 into the pET21(+) vector.
Figure 1. Domain diagram of His-RIG-I-CARDapaf1

Test

During a discussion with Dr. Fujita as part of our Human Practices activities, we learned that expressing and purifying RIG-I in E. coli is technically challenging. As a result, researchers often use alternative expression systems such as insect cells.

Learn

We realized that using standard methods to express RIG-I in E. coli made purification difficult, and that a modified approach would be required to obtain functional protein.
Cycle2

Design

To improve the expression and solubility of RIG-I-CARDapaf1 in E. coli, we searched for prior research and found that fusion with solubility-enhancing tags such as SUMO could be effective. Based on this, we redesigned the plasmid by adding an N-terminal SUMO tag to the construct.
Figure 2. Domain diagram of SUMO-tagged RIG-I-CARDapaf1

Build

We constructed the redesigned plasmid and transformed it into E. coli for protein expression and purification.

Test

We assessed protein expression and purification by SDS-PAGE. As a result, we were able to purify a small but detectable amount of SUMO-tagged RIG-I-CARDapaf1.
Figure 3. SDS-PAGE of purified RIG-I-CARDapaf1.

Learn

We learned that appending a SUMO tag to the N-terminus of RIG-I-CARDapaf1 improves its solubility and allows partial purification in E. coli. This confirmed the utility of fusion tags in improving expression of difficult-to-purify proteins.
Cycle3

Design

To evaluate the dsRNA-binding properties of recombinant RIG-I and its mutants, we needed to prepare a well-defined double-stranded RNA (dsRNA) substrate. Although many studies on RIG-I use synthetic analogs like poly(I:C), such molecules are unsuitable for assays requiring sequence-specific quantification—such as pull-down assays followed by qRT-PCR. Therefore, inspired by prior studies [e.g., Cell (2019), https://doi.org/10.1016/j.cell.2019.03.004], we designed a method to synthesize dsRNA in vitro from a DNA template using the MEGAscript® T7 Transcription Kit.
Figure 4. DNA templates used for in vitro transcription.
Two templates with different T7 promoter placements were designed to produce complementary single-stranded RNAs.

Build

We ordered DNA fragments (gBlocks) from IDT, designed with opposing T7 promoters to transcribe complementary RNA strands. These templates were then used with the MEGAscript® Kit to synthesize RNA strands. We transcribed two complementary ssRNAs simultaneously in a single tube, allowing them to anneal naturally during the transcription reaction and form dsRNA.

Test

We transcribed RNA from each template and purified the products using the PureLink® RNA Mini Kit. The purified RNAs were analyzed by polyacrylamide gel electrophoresis (PAGE) to confirm successful formation of dsRNA.
Figure 5. Gel electrophoresis of synthesized dsRNA showing the expected size band.

Learn

We successfully synthesized dsRNA of the expected length and confirmed its formation through PAGE. This provided us with a suitable reagent for downstream functional assays, such as verifying RNA binding by the recombinant RIG-I fragments.
Cycle4

Design

To investigate whether the purified RIG-I-CARDapaf1 protein could bind to dsRNA in vitro, we planned a pull-down assay. The experiment involved incubating purified His-tagged RIG-I-CARDapaf1 with in vitro-transcribed dsRNA in binding buffer, then using nickel beads to capture the His-tagged protein along with any bound RNA. We aimed to detect and quantify the pulled-down dsRNA to assess binding efficiency.

Build

To validate binding specificity, we prepared three experimental conditions: no RIG-I-apaf1, with RIG-I-apaf1, and with a mutant RIG-I-apaf1. The RNA pulled down under each condition was then purified for further analysis.

Test

We quantified the amount of RNA recovered from each condition using qRT-PCR. The percent recovery was calculated relative to the initial input RNA. Each condition was performed in triplicate (n = 3), and results were statistically analyzed.
Figure 6. Results of the dsRNA pull-down assay using RIG-I-CARDapaf1.

Learn

Surprisingly, a substantial amount of dsRNA was recovered even in the no-protein control, indicating strong non-specific binding to the nickel beads. As a result, we could not clearly distinguish between protein-dependent and background binding. We suspect that insufficient washing steps during bead handling contributed to the high background. Moving forward, we plan to optimize the washing conditions and potentially use alternative affinity tags or blocking agents to reduce non-specific binding. With these improvements, we anticipate that dsRNA binding by RIG-I-CARDapaf1 can be visualized more clearly in future assays.
References

Apoptosis Inducing Module

Initial Plan
COCCO is a chimeric protein designed by replacing the CARD domain of RIG-I (CARDrigI) with the CARD domain of Apaf1 (CARDapaf1). This modification is intended to enable COCCO to interact with the CARD domain of Caspase-9 (CARDcasp9), thereby triggering apoptosis. To validate this interaction, we planned to purify CARDcasp9 and test its binding with the RIG-I-CARDapaf1 construct previously described in our experiments. Specifically, we designed a pull-down assay in which bead-immobilized CARDcasp9 would be used to capture RIG-I-CARDapaf1 from solution.
Cycle1

Design

To enable CARDcasp9 to bind to Glutathione Sepharose, we designed GST-CARDcasp9, which features a GST-tag sequence from pGEX-6p-1 at its N-terminus. For expression in E. coli, the codon-optimized nucleotide sequence for GST-CARDcasp9 was inserted into the pET21(+) vector.

Build

The plasmid was synthesized by Twist Bioscience. We transformed the prepared plasmid into E. coli BL21(DE3) to express the protein.

Test

When the OD of the culture reached 0.4, the culture was induced under the following one of two conditions:
  • Incubation with 1 mM IPTG at 37°C for 1.5 hours
  • Incubation with 1 mM IPTG at 16°C overnight (o/n)
Figure 7. SDS-PAGE analysis of GST-CARDcasp9 expression Cycle1

Learn

Since we could not obtain the expected amount of recombinant protein under multiple culture conditions, we suspected there might be a problem with the plasmid design. Upon investigation, we found that although the amino acid sequence of the GST tag was correct, its nucleotide sequence was different from the original pGEX-6P-1 vector due to codon optimization. We learned that the nucleotide sequence itself, not just the amino acid sequence of the GST portion, can have a critical impact on protein expression.
Cycle2

Design

Instead of using the construct created with pET21(+), we decided to express GST-CARDcasp9 using the original pGEX-6P-1 vector, which contains the native nucleotide sequence for the GST tag. Our hypothesis was that using the native GST coding sequence would improve protein expression efficiency.

Build

We excised the CARDcasp9 fragment from the plasmid generated in Cycle 1 using restriction enzyme digestion. This fragment was then ligated into a pGEX-6P-1 vector that had been digested with the same enzymes, resulting in a new plasmid construct. We transformed this plasmid into E. coli BL21(DE3) and proceeded with protein expression.

Test

The culture was induced with 1 mM IPTG at 37°C for 1.5 hours. Recombinant protein expression was assessed via SDS-PAGE.
Figure 8. SDS-PAGE analysis of GST-CARDcasp9 expression Cycle2

Learn

By switching to the pGEX-6P-1 vector with the native GST sequence, we were able to purify a significantly greater amount of recombinant protein. This result highlighted that even when both the amino acid sequence and the promoter remain unchanged, variations in the nucleotide sequence of the open reading frame (ORF)—particularly within regions encoding affinity tags—can dramatically influence the expression yield of recombinant proteins in E. coli.
Cycle3

Design

Building upon our successful preparation of GST-CARDcasp9 immobilized on Glutathione Sepharose beads, we next designed a pull-down assay to test for interaction with RIG-I-CARDapaf1. To facilitate detection, we inserted a FLAG tag into the RIG-I-CARDapaf1 construct. As a positive control, we planned to test FLAG-tagged CARDapaf1 alone, which was expected to interact with CARDcasp9 based on previous literature.

Build

We constructed a plasmid encoding FLAG-CARDapaf1 and expressed it in E. coli BL21(DE3). The recombinant protein was purified and its expression confirmed by SDS-PAGE.

Test

Pull-down assays were performed using Glutathione Sepharose beads bound to GST-CARDcasp9. Proteins that co-precipitated were detected via Western blotting using an anti-FLAG antibody. However, no interaction was observed: neither RIG-I-CARDapaf1 nor FLAG-CARDapaf1 was pulled down.
Figure 9. CARDapaf1-CARDcasp9 interaction assay.
Western blot analysis shows that neither RIG-I-CARDapaf1 nor FLAG-CARDapaf1 was co-precipitated by GST-CARDcasp9.

Learn

Although a direct in vitro interaction between CARDapaf1 and CARDcasp9 has been previously reported by Tamura et al. (2018) [https://www.nature.com/articles/s41598-018-19845-6], we were unable to replicate this result under our experimental conditions. This suggests that the assay conditions—such as buffer composition, protein concentrations, or wash stringency—may not have been optimal. We concluded that further optimization will be necessary to visualize the interaction between these CARD domains.
References

Integration of Two Modules

Initial Plan
Our aim was to induce apoptosis through the activation of Caspase-9 by achieving two conditions in our engineered constructs, RIG-I-CARDapaf1 and PKR-CARDapaf1: Binding of double-stranded RNA (dsRNA) via their respective dsRNA-binding domains and Interaction between the CARD domain of Apaf1 and the CARD domain of Caspase-9. To validate that both conditions are met within living cells, we planned to create human or chicken cells stably or transiently expressing these constructs. Upon transfection with dsRNA, we would assess whether apoptosis was successfully induced, thereby verifying the functionality of the integrated module in a cellular context.
Cycle1

Design

We decided to construct plasmids for transfection into HEK293 and DF-1 cells in order to express our target proteins. Since COCCO is designed to induce apoptosis via oligomerization, we were concerned that overexpression might lead to unintended oligomer formation and premature apoptosis. Therefore, we considered promoter strength optimization to be a critical step.

Build

Using the CMV promoter, we designed plasmids expressing the following genes:
  • RIG-I-CARDapaf1
  • PKR-CARDapaf1
  • PKR-ΔCaspase9
  • non CARD RIG-I
  • CARDapaf1
  • PKR
  • ΔCaspase9
  • Empty vector (control)

Test

We co-transfected these plasmids with pCMV-DsRed into both HEK293 and DF-1 cells. Expression was verified via DsRed fluorescence, and we monitored cells under a microscope to check for signs of apoptosis potentially caused by unintended COCCO activity.
Figure 10. Microscopy images of transfected cells.
The top left image expresses HEK293 expressing RIG-I-CARDapaf1, and the top right one expresses HEK293 expressing DsRed. Also, the bottom left one expresses DF-1 expressing RIG-I-CARDapaf1, and the right bottom one expresses DF-1 expressing DsRed

Learn

The successful expression of DsRed confirmed that the transfection and promoter strength were adequate. Moreover, the absence of spontaneous apoptosis in cells expressing COCCO indicated that oligomerization was not occurring inappropriately. From this cycle, we concluded that the CMV promoter provides a suitable expression level for our experiments in both HEK293 and DF-1 cells.
Cycle2

Design

To integrate the apoptosis-inducing module in mammalian cells, efficient transfection of plasmid DNA is essential. Since our experiments involved two different cell lines—HEK293 (human) and DF-1 (avian)—we recognized the importance of selecting an appropriate transfection method tailored to each. We consulted Dr. Shinnosuke Honda regarding transfection methods suitable for HEK293 and DF-1 cells. Based on his recommendation and prior experience, we chose to use Lipofectamine™ 2000 Reagent for plasmid transfection.

Build

Following the official Lipofectamine™ 2000 protocol [¹], we transfected our previously constructed plasmids into HEK293 and DF-1 cells. In parallel, we co-transfected pCMV-DsRed to assess transfection efficiency via fluorescence.

Test

To evaluate the transfection efficiency, we observed DsRed fluorescence in HEK293 and DF-1 cells under a fluorescence microscope.
Figure 11. Fluorescence microscopy of cells transfected with pCMV-DsRed
The left image expresses HEK293 expressing DsRed, and the right one expresses DF-1 expressing DsRed.

Learn

From the fluorescence images, we observed strong DsRed expression in HEK293 cells, indicating successful transfection and protein expression. However, DF-1 cells exhibited weak or minimal fluorescence. This result suggests that, although plasmid transfection using Lipofectamine™ 2000 is possible in DF-1 cells, it does not lead to sufficient protein expression in this cell line. Therefore, alternative transfection reagents or methods may be necessary for effective expression in DF-1 cells.
Cycle3

Design

Since the results from Cycle 2 indicated low transfection efficiency in DF-1 cells using Lipofectamine™ 2000, we sought to optimize the protocol specifically for this cell line. Enhancing plasmid delivery efficiency was critical for reliably evaluating downstream apoptosis assays. Based on the observations from Cycle 2, we decided to explore an alternative transfection reagent to improve protein expression in DF-1 cells. Specifically, we compared the transfection efficiencies of Lipofectamine™ 2000 Reagent and FuGENE® 6 Reagent using pCMV-DsRed as a reporter.

Build

We transfected DF-1 cells with pCMV-DsRed using both Lipofectamine™ 2000 and FuGENE® 6, following each manufacturer's recommended protocol [¹][²].

Test

To assess transfection efficiency, we observed the fluorescence intensity of DsRed under a fluorescence microscope in DF-1 cells transfected with each reagent.
Figure 12. Comparison of transfection efficiency between Lipofectamine2000 reagent and FuGENE6 reagent
The left image expresses DF-1 transfected with DsRed using Lipofectamine, and the right one expresses DF-1 transfected with DsRed using FuGENE.

Learn

The fluorescence images showed markedly stronger DsRed expression in DF-1 cells transfected with FuGENE® 6 compared to Lipofectamine™ 2000, indicating improved transfection efficiency. Based on these results, we concluded that FuGENE® 6 is a more suitable transfection reagent for DF-1 cells and will use it in future experiments involving this cell line.
References

Dry Lab Cycle

Overview

To utilize modeling in our project, we performed a viral transmission simulation, which cannot be observed through experiments, and a protein simulation that could also integrate evaluation by the Wet Lab.

Viral Transmission Modeling

Cycle1

Design

To prove the effectiveness of COCCO, we divided the viral transmission into three levels: intracellular, intercellular, and inter-individual. Therefore, we set the objective of our initial viral transmission modeling to prove the following points using a mathematical model.
  • Within a cell, COCCO causes the host cell to die faster than the virus can proliferate.
  • COCCO suppresses viral spread among a cell population, preventing the virus from proliferating to the extent that it can be transmitted to other individuals.
  • The spread of viral infection is suppressed within a population of chickens equipped with COCCO.

Build

At the intracellular level, we analyzed the apoptosis reaction network and simulate that the induction of apoptosis is sufficiently rapid. At the intercellular level, we simulated that increasing the cell death rate suppresses viral proliferation. At the inter-individual level, we simulated that the spread of infection is suppressed because even if infected, individuals transition from an eclipse phase to a susceptible state without becoming infectious. Furthermore, we connected the simulations of these three levels to simulate that the spread of viral infection is suppressed within a chicken coop of COCCO-equipped chickens.

Test

After reviewing several databases related to metabolic and signaling pathways, we found a lack of detailed information on the apoptosis pathway in chickens, which limited the reliability of intracellular-level modeling.

Learn

We learned that the chicken apoptosis pathway is not well understood. Therefore, we concluded that the intracellular dynamics modeling could not be sufficiently reliable.
Cycle2

Design

We decided to exclude the intracellular dynamics modeling and perform a two-level viral transmission simulation for the intercellular and inter-individual levels.

Build

We decided to use the most widely used TIV model and SIR model to describe the intercellular and inter-individual viral transmission, respectively. We also determined how to proceed with Human Practices to receive advice on necessary assumptions and other points of caution when handling these models.

Test

In Human Practices (HP link), it was pointed out that the inter-individual viral transmission modeling would not be a reliable simulation unless we used data obtained from our own observations.

Learn

We realized that we could make almost no claims from the inter-individual viral transmission simulation, as we cannot collect observational data on the spread of viral infection in COCCO-equipped chickens or chicken coops.
Cycle3

Design

We decided to exclude the inter-individual dynamics modeling and perform a viral transmission simulation at the intercellular level.

Build

We believed that the time elapsed since a virus entered a cell was important for the simulation of COCCO. Therefore, we introduced the quantity of time known as "infection age" to represent this elapsed time and decided to use an age-structured TIV model[1].

Test

In Human Practices (HP link), it was pointed out that we should construct a minimal model for our claims and that we should simulate how the infection dynamics change when the apoptosis intensity is varied.

Learn

We considered the age-structured TIV model to be a minimal model for our purposes and thought that a meaningful modeling approach would be to evaluate the strength of the effect of a decreased death rate on the value that determines the behavior of the infection dynamics.

Protein Optimization

Cycle1

Design

To prevent the fusion protein, in which the CARDs of RIG-I were replaced with the CARD of APAF1, from always remaining in an active state that can transmit signals, we decided to predict the optimal amino acid mutations to achieve the ideal function of COCCO through in silico protein simulation.

Build

We investigated available software and constructed a workflow EVOLVE (version 1) to search for and select the optimal mutant amino acid sequence in the following five steps.
  • Mutable Site Selection
  • Mutation Introduction
  • Structure Prediction
  • Visual Inspection
  • Docking & Stability Analysis

Test

We executed this EVOLVE (version 1) and selected five sequences each for the mutated Helicase2i domain and APAF1 CARD domain. The results can be found on the Model page under Protein Optimization.

Learn

We learned that the workflow allows for the in silico optimization of the amino acid sequences of protein domains to have the functions required for COCCO.
Cycle2

Design

Based on advice from Human Practices, we learned about the usefulness of a software called ProteinMPNN[2], which predicts amino acid sequences based on 3D structures, and decided to use it.

Build

We devised EVOLVE (version 2) that uses ProteinMPNN as the software in Step 2: Mutation Introduction.

Test

We executed this workflow and generated two amino acid sequences with mutations introduced into the Helicase2i domain. The results can be found on the Model page under Protein Optimization.

Learn

We found that EVOLVE (version 2) could similarly be used to design protein domains with the optimized amino acids required for COCCO.
References
    [1]Nelson, P. W., Gilchrist, M. A., Coombs, D., Hyman, J. M., & Perelson, A. S. (2004). An Age-Structured Model of HIV Infection that Allows for Variations in the Production Rate of Viral Particles and the Death Rate of Productively Infected Cells. Mathematical Biosciences & Engineering, 1(2), 267–288.

    https://doi.org/10.3934/mbe.2004.1.267

    [2]Dauparas, J., Anishchenko, I., Bennett, N., Bai, H., Ragotte, R. J., Milles, L. F., Wicky, B. I. M., Courbet, A., De Haas, R. J., Bethel, N., Leung, P. J. Y., Huddy, T. F., Pellock, S., Tischer, D., Chan, F., Koepnick, B., Nguyen, H., Kang, A., Sankaran, B., . . . Baker, D. (2022). Robust deep learning–based protein sequence design using ProteinMPNN. Science, 378(6615), 49–56.

    https://doi.org/10.1126/science.add2187