Overall Design

Aim 1: Identification of cytokine profiles correlated to cytokine storm in patients

  • Research into which cytokines are most prominent in the COVID-19 cytokine storm
  • Looking into downstream signaling of cytokines to find appropriate biomarkers
  • Building a circuit to sense our chosen biomarkers

Aim 2: Identification of anti-inflammatory output protein as a response to the cytokine storm

  • Research into cytokine blockers, receptor antagonists, scFv’s
  • Research of current treatments for cytokine storms and hyperinflammation

Aim 3: Design iterations for processing our cytokine inputs

  • Cascades system involving two receptors with distinct inputs
  • Implementing existing analog circuitry in bacterial cells into a mammalian system
  • PERSIST-based system, using endoribonucleases at the translational level

Aim 1

FIGURE 1: The ROC curve of plasma cytokine levels on admission for patients with and without ARDS during hospitalization. The ROC curve was calculated using SPSS. The AUC of the ROC curve for IL-1ra, IL-6, HGF, MCP-3, M-CSF, and IP-10 was estimated using the expression levels upon admission. A, The ROC curve of each cytokine. B, The ROC curves of different combinations of IP-10, MCP-3, and IL-1ra. All the P values were less than .05. (Yang et al.)

Forty-eight cytokines in the plasma samples of 50 COVID-19 patients, including 11 critically ill, 25 severe, and 14 moderate patients, were measured and analyzed in combination with clinical data (Yang et al.). Based on the receiver operating characteristic (ROC) curves by Yang et al., IP-10 and MCP-3, two cytokines secreted by the IFN-γ cascade, when present together in elevated concentrations, served as promising biomarkers for severe COVID-19 cytokine storm (CCS).

FIGURE 2: Downstream signaling pathways of the cytokines IP-10 and MCP-3. (Adapted from Creative Diagnostics and Semantic Scholar)

IP-10 and MCP-3 canonical signaling cascades result in increased expression of ELK-1 and NF-κB, respectively (Figure 2). Thus, we identified ELK-1 and NF-κB as two transcription factors serving as biomarkers of the CCS. Although the other transcription factors shown in Figure 2 (e.g. ATF2 for IP-10 and AP-1 or CREB for MCP-3) may be equally valid biomarkers of the CCS, we decided to use ELK-1 and NF-κB as the inputs for our computationally-designed sensing circuit. In the future when we have lab access, we hope to iterate through various sensing circuitry using combinations of the different downstream transcription factors, and compare each circuit’s efficacy in reducing CCS and its associated inflammation.

Aim 2

We considered various options for anti-inflammatory actuation targeting cytokines integrated in the CCS to prevent native receptor signal transduction, including: competitive agonists, receptor antagonists, and soluble decoy receptors.

Antagonists and decoy receptors were potential candidates to alleviate the effects of the cytokine storm because they both decrease signaling of receptor ligands, in this case our cytokines. However, currently known receptor antagonists are limited to the IL-1 family of cytokines only, and an issue we found when trying to apply decoy receptors to our circuit is that decoy receptors often are unable to keep their structure when transmembrane regions are removed.

Thus, we decided to avoid decoy receptors and receptor antagonists in designing our anti-inflammatory actuation protein.

Instead we chose to use scFvs (single chain variable fragments), which are a fusion of light and heavy chain fragments capable of binding to a target ligand, connected by a peptide linker. scFvs are more specific and stable than decoy receptors and are available for a wide variety of cytokines, rather than just one. Based on the specificity and efficacy of drugs like tocilizumab (IL-6 antibody) for alleviating overactive immune disorders, we implemented scFvs as the AND gate output, due to their ease of translation and structural integrity. Additionally, as some of the whole antibodies against cytokine targets in our CCS system have not been clinically well-characterized in humans, scFvs reduced immunogenicity was appealing in our system’s context.

Aim 3

Our team was interested in AND gates because we wanted to respond to the presence of both of our chosen CCS biomarkers. We iterated through three versions of AND logic for the two primary cytokine biomarkers: firstly, cascades (Schukur et al.), secondly, logarithmic-scale (Daniel et al.), and lastly, a PERSIST-based system (DiAndreth et al.). Schukur et al.’s cascade method employs two receptors with distinct inputs: upon binding of the first input to the receptor, an intermediate necessary for the second transduction pathway is produced; this second pathway eventually transcribes an output gene. Daniel et al.’s synthetic analog gene circuits produced through variable copy number AraC-based plasmids were shown to logarithmically transform a wide dynamic range of input inducer concentrations into output protein levels. However, when trying to apply this analog circuit design to mammalian cells, we ran into difficulties finding a system mechanistically analogous to the bacterial AraC protein.

Programmable Endonucleolytic Scission-Induced Stability Tuning (PERSIST), a CRISPR-based RNA regulation system developed in the Weiss lab that can produce switchlike “ON” and “OFF” responses (Diandreth et al.), offered promising tunable behavior for our circuitry with digital specificity at the post-transcriptional mRNA level.

FIGURE 3: Full network circuit design, responsive to COVID-19 cytokine storm biomarkers’ (IP-10 + MCP-3) nuclear transcription factors (ELK-1 + NF-κB) , using digital PERSIST “OFF” logic (CasE, Csy4, Cas6), and producing an individualized anti-inflammatory, single chain antibody fragment output (scFv).

To summarize the full circuit depicted in Figure 3, we will sense the cytokines IP-10 and MCP-3 which correspond to the transcription factors ELK-1 and NF-κB respectively. CasE is expressed upon activation by ELK-1, and Csy4 is expressed upon activation by NF-kB. Using the CMV promoter, Cas6 is constitutively expressed but when either or both CasE and Csy4 are present, the mRNA for Cas6 will be cut upstream of the gene sequence, so Cas6 expression will decrease. For our output fusion protein, we have another plasmid with the constitutive promoter CMV, and likewise there is an ERN cut site upstream of the gene sequence. Thus, when both IP-10 and MCP-3 are present, scFv will be produced, inhibiting our cytokine of interest.

We developed an AND gate with five PERSIST-based plasmids. IP-10 and MCP-3 signal downstream phosphorylation pathways to activate nuclear transcription factors (nTFs) ELK-1 and NF-κB, respectively. Two minimal constitutive promoters (here, CytoMegaloVirus, CMV) are then induced by nTF-responsive element binding, upstream of ERN genes. In choosing our ERN components for translation, we evaluated the sensitivity with the relative abundance of the biomarker (nTFs ELK-1 and NF-κB) in plasma. In CCS, MCP-3 has been observed about 103 fold lower than IP-10; Csy4, an experimentally stronger “OFF” ERN, was thus placed downstream of the NF-KB-responsive promoter, while CasE, a slightly weaker ERN, was used for the IP-10 sginaling pathway. Cas6, the weakest ERN of the Cas family evaluated, when produced by either of two CMV-constitutively-expressed transcripts, degrades the scFv mRNA transcripts and causes decreased translation of anti-inflammatory soluble scFv with a slight buffer to allow for IP-10 and MCP-3 elevation to be reached. To further enhance the system dynamics by decreasing the time required to reach steady state, a PEST degradation tag can be fused to the Cas6 sequence. Our AND gate is thus engineered to produce a dose-responsive output when there is adequate IP-10 and MCP-3 to result in the degradation of Cas6 transcripts.


Daniel, R., Rubens, J., Sarpeshkar, R., & Lu, T. K. (2013). Synthetic analog computation in living cells. Nature 497, 619–623 doi:10.1038/nature12148

DiAndreth, B., Wauford, N., Hu, E., Palacios, S., & Weiss R. (2019). PERSIST: A programmable RNA regulation platform using CRISPR endoRNases. bioRxiv, 12(15) 1-17. doi:10.1101/2019.12.15.867150

Schukur, L., Geering, B., Charpin-El Hamri, G., & Fussenegger, M. (2015). Implantable synthetic cytokine converter cells with AND-gate logic treat experimental psoriasis. Science Translational Medicine, 7(318), 318ra201. doi:10.1126/scitranslmed.aac4964

Yang, Y., Shen, C., Li, J., Yuan, J., Wei, J., Huang, F., Wang, F., Li, G., Li, Y., Xing, L., Peng, L., Yang, M., Cao, M., Zheng, H., Wu, W., Zou, R., Li, D., Xu, Z., Wang, H., Zhang, M., … Liu, Y. (2020). Plasma IP-10 and MCP-3 levels are highly associated with disease severity and predict the progression of COVID-19. The Journal of allergy and clinical immunology, 146(1), 119–127. doi:10.1016/j.jaci.2020.04.027

Here is our presentation for our weekly update meetings (edited for length)!

This page was written by Erin Shin