Engineering Overview

Our Wet Lab Team aimed to design, construct, and validate a modular, programmable RNA-sensing platform using ADAR editing in living cells. By coupling programmable RNA sensors to split protein outputs, our system enables the detection of disease-specific RNA signatures and the activation of therapeutic proteins only when multiple targets are simultaneously present.

We focused on hepatitis B virus–induced hepatocellular carcinoma (HBV–HCC) as a clinically relevant model, targeting the HBx (viral) and GPC3 / AKR1B10 (host) transcripts. We conducted multiple Design–Build–Test–Learn cycles to iteratively refine sensor design, validate logic gating, and prototype therapeutic outputs.

Cycle 1: Designing and Validating the HBx Viral Sensor

Design

Our first goal was to create a RADAR sensor that could selectively recognize HBx mRNA, a viral transcript critical for HBV replication and oncogenic transformation. Using the Kaseniit et al. (2022) RADAR algorithm, we designed sensor-trigger pairs targeting conserved regions of the HBx transcript.

We performed a multiple-sequence alignment of HBV genomes and identified the region between AUG1 and AUG2 as both highly conserved across HBV genotypes and accessible within the mRNA secondary structure. Two sensor variants were designed:

HBx_Broad Sensor: Targeting conserved regions to detect diverse HBV strains.

HBx_CellLine Sensor: Optimized for Hep3B-specific HBV transcripts. Each sensor was paired with a complementary trigger RNA, and cloned downstream of a constitutive promoter in a mammalian expression vector.

Each construct contains a RADAR recognition domain that forms a short double-stranded RNA with the target transcript, recruiting endogenous ADAR enzymes to deaminate a specific adenosine within a synthetic UAG stop codon embedded in the GFP coding sequence. Successful editing converts the stop codon to tryptophan, restoring translation and producing GFP fluorescence upon ADAR-mediated editing and translation reactivation.

We predicted RNA accessibility and folding using RNAfold to identify energetically favorable regions within HBx that maximize base pairing while minimizing intramolecular secondary structure.

Build

The two sensor variants were synthesized as gBlocks using clonal Twist synthesis and cloned into a mammalian expression vector downstream of a CMV promoter using In-Fusion cloning and transformed them into E. coli DH5α for amplification. Each construct contained:

• mCherry reporter
• 3' UTR sensor sequence (complementary to HBx target) containing the RADAR-embedded stop codon
• GFP reporter output

Plasmids were sequence-verified using Sanger sequencing and transfected into HEK293T and Hep3B cells using Lipofectamine 3000.

Test

HEK293T cells were transfected with either HBx_Broad or HBx_Hep3B sensors, along with their matching or scrambled trigger plasmids, using Lipofectamine 3000. Transfection efficiency was optimized by varying cell confluency (70–90%), DNA:Lipofectamine ratios, and incubation time. Both HBx sensor and trigger plasmids were co-transfected at equimolar ratios.

Specificity testing was performed by introducing a 5% sequence mismatch in the trigger RNA. Despite this mismatch, both sensors maintained a clear signal with minimal off-target activation, confirming tolerance to viral mutation variability.

HBx_Broad Sensor:

One A-C editable mismatch between sensor and Broad Trigger designed, as predicted.

HBx_CellLine Sensor:

One A-C editable mismatch between the sensor and Broad Trigger is designed in addition to 7 mismatched bases. Used to test functionality and specificity even in the case of a 5% mismatch due to frequent viral sequence mutations and variances, as predicted.

After 48 hours, fluorescence was measured using flow cytometry and fluorescence microscopy. Cells transfected with HBx_Broad + trigger or HBx_CellLine + trigger showed strong GFP expression compared to control cells lacking the trigger, validating the specificity of detection. White signals indicate GFP-positive cells; red signals represent sensor presence. Even in the edge case of a 5% mismatch between the broad trigger and the cell line sensor, we still observed measurable GFP output, indicating that common viral mutations do not significantly alter sensor performance.

Flow cytometry analysis of GFP fluorescence in HEK293T cells transfected with the HBx_Broad and HBx_Hep3B sensors in response to matched and mismatched synthetic triggers indicates a 13-fold and 23-fold, respectively, increase in GFP output relative to the mismatched (scramble) controls. Complete match triggers also produced a statistically significant increase in fluorescent intensity compared to partial match triggers (p=0.0233). Results confirm both the sensitivity and ability to tolerate limited sequence variation in viral sequences.

Single replicate quantitative evaluation of HBx_Broad vs. HBx3b sensor performance.

Learn

From this cycle, we learned that:

1. The RADAR system successfully detects viral RNA in mammalian cells.
2. The Broad sensor design offers superior performance and cross-genotype stability.
3. Moderate sequence mismatches (≤5%) do not abrogate detection, ensuring real-world applicability in diverse HBV variants. These insights established a robust viral sensor for integration into future multi-input circuits.

Cycle 2: Host RNA Sensor (GPC3 and AKR1B10)

Design

As we continued to research the trigger sequences we recognized that low expression of viral transcripts may not allow for HBx to be a good input so we looked for endogenous mRNAs associated with HCC. To achieve selective recognition of cancerous over healthy hepatocytes, we next designed sensors for GPC3 (Glypican-3) and AKR1B10 (Aldo-Keto Reductase Family 1 Member B10), two transcripts consistently upregulated in HBV–related HCC relative to healthy liver tissue, to enable combinatorial disease detection.

Both are expected to have a singular editable A-C mismatch near the center of the sequence. Trigger plasmids were synthesized using clonal twist synthesis. Each plasmid contained the reverse complementary sequence of the respective sensor sequence under a constitutive promoter (CMV), followed by mTagBFP to mark the presence of the trigger in relation to edits made by the ADAR enzyme 24 hours following transfection.

GPC3:

AKR1B1O:

Using GEPIA2 and DepMap transcriptomic datasets, the Dry Lab Team identified candidate transcripts, filtering by:

• High tumor-to-normal log₂ fold change (>2)
• Minimal basal expression in normal liver (TPM < 1)
• Presence in the Hep3B transcriptome

The RADAR algorithm was used to generate sequence-specific sensor regions complementary to accessible mRNA domains predicted via RNAplfold.

Build

Each host sensor was cloned using the same RADAR-GFP backbone used in Cycle 1 as the viral sensor, enabling direct comparability of performance. Constructs were sequence-verified via Sanger sequencing, and plasmids were co-transfected into HEK293T and Hep3B cells, either individually or in combination with viral triggers to test potential crosstalk.

Test

HEK293T cells were co-transfected with each sensor and its corresponding trigger or scrambled control. After 48 hours, GFP expression was assessed by microscopy and flow cytometry.

Both GPC3 and AKR1B10 sensors yielded significant activation when co-transfected with their respective triggers compared to the scramble condition, confirming the specificity of our sensor system.

Mean calculations of our flow cytometry values revealed that our AKR1B10 sensor exhibited a 20-fold increase in GFP fluorescence in the top 1% of sensor expressing cells (p = 0.08 following t-test) whereas GPC3 sensor exhibited a 15-fold increase in GFP fluorescence (p = 0.00262 following t-test). The sensor imaging show mCherry reporter signalling the presence of the sensor, while the output editing columns show white filling in areas where a sensor-trigger match resulted in a GFP output.

Single replicate quantitative evaluation of AKR1B10 sensor performance.

Single replicate quantitative evaluation of GPC3 sensor performance.

Learn

We confirmed that the RADAR system can be modularly extended to detect case-specific host mRNAs, which enabled us to proceed to combinatorial logic-gate design using validated sensors as interchangeable inputs. Each sensor maintained high specificity and orthogonality, validating the platform's capacity for sensing combinatorial inputs. We also observed that GPC3 produced a slightly higher signal-to-background ratio in HCC-related lines, which informed its selection for downstream AND-gate integration.

Cycle 3: Logic Gating with Split GFP (AND Gate)

Design

Having validated both viral and host sensors, we sought to integrate them into a dual-input AND logic gate, ensuring signal activation only in cells expressing both HBx and GPC3- characteristic of HBV-transformed hepatocytes. We used a split GFP design, where:

• Sensor 1 (HBx) controls the expression of the GFP1–10 fragment.
• Sensor 2 (GPC3) controls the expression of the GFP11 fragment.
• Fluorescence should occur only when both fragments are translated and properly folded, serving as a molecular AND gate proof of concept.

Build

Each sensor was fused to its respective split GFP fragment and cloned into mammalian expression plasmids. Before integration into the AND-gate system, the split GFP halves were tested under constitutive promoters to confirm proper reconstitution (positive control). Co-transfections were performed in HEK293T cells in 48-well format, including:

1. GFP1–10 alone
2. GFP11 alone
3. Both halves together (positive control)
4. AND-gated sensor circuits with all trigger combinations:
1. Both triggers (dual match)
2. HBx trigger only
3. GPC3 trigger only
4. Scramble (no trigger-sensor matches)

Test

Constitutive expression confirmed split GFP reconstitution (fluorescent signal only when both halves were co-expressed). No fluorescence was observed with either half alone, while a measurable GFP signal was detected only when both fragments were co-expressed. This confirms proper reassembly of the split GFP protein for the AND-gated sensor to be tested.

HEK293T cells were transfected with GPC3 + HBx AND-gated sensors and co-transfected with either both GPC3 and HBx_Broad triggers (dual match), just one of them, or neither of them (scramble). Fluorescence microscopy revealed that GFP signal appeared only when both HBx and GPC3 triggers were co-expressed. Single-input conditions (only HBx or only GPC3 trigger) produced minimal signal, confirming logical AND behavior.

These results are corroborated by flow cytometry analysis, which indicated that complete match triggers relative to both sensors produced the highest GFP output compared to mismatched inputs, confirming the high specificity of our system. However, these are preliminary results and would require more replicates to calculate statistical significance. These results show proper AND logic behavior with minimal leakiness.

Quantitative evaluation of AND-Gated sensor performance.

Learn

These experiments validated that multiple RNA sensors can be combined into higher-order logic circuits, enabling precise multi-input control of gene expression. It also demonstrated the scalability of the RADAR framework for complex biological computation within mammalian cells. The AND gate ensures high therapeutic expression is confined to cells where both HBx and GPC3 transcripts are present, allowing for disease-specific activation for our application. Although there appeared to still be some GFP expression in the single trigger match cases, the 2-fold increase in output expression in the dual match case proves our system's potential and validity. Leakiness and tunability are to be further optimized in experiments to come.

Cycle 4: Split IL-2 Output (In Progress)

Design

Implementing our proof-of-concept towards a therapeutic application, we designed a split IL-2, an immunostimulatory cytokine used in cancer immunotherapy, as a therapeutic output to replace the split GFP. The goal was to create a system where IL-2 is secreted only when both viral (HBx) and host (GPC3) transcripts are detected.

Build

We are in the process of cloning the N- and C-terminal fragments of IL-2 downstream of the HBx and GPC3 sensors, respectively, preserving the same stop-codon editing framework. Each half will be fused to a signal peptide (IL2sp) to direct them into the secretory pathway, where reconstitution can occur following entry into the endoplasmic reticulum (ER). All constructs will be sequence-verified, transfected and expressed in HEK293T cells in the same way as our split GFP experiments.