Week 1 Lab: Pipetting

Week 2 Lab: DNA Gel Art

Week 3 Lab: Opentrons Art

Week 6 Lab: Gibson Assemly

Week 6: Gibson Assembly

Group members: Louisa Zhu, Shitong, Jasmin

Part 1

1. PCR

Figure 1. PCR reaction setup tables for the Backbone DNA Fragment (top) and Color DNA Fragment (bottom), including reagent volumes for a 25 µL total reaction.

We then ran the PCR reaction with the following thermocycler settings:

Figure 2. Thermocycler programs for the backbone PCR (BB_PCR, left) and color fragment PCR (COLR_PCR, right). Both protocols use 26 cycles with a final extension at 72°C for 5 minutes.

2. Gel Eletrophoresis

Protocol credit to Louisa:

1. Take 2 µL of each mixture and transfer into new labeled PCR tubes
2. Pipette 2 µL of mUAV into a new tube
3. Add 20 µL of water to each PCR tube
4. Unpack gel electrophoresis cassette and load into machine
5. Pipette DNA Ladder into first well
6. Pipette 20 µL of mixture from each new PCR tube into correct wells (6 full wells total)
7. Use the automatic setting for 1%, wait 10 minutes

Figure 3. Agarose 1% gel electrophoresis result showing PCR products. The DNA ladder (M) is in the first lane. Bands are visible for the backbone and color fragments, confirming successful amplification.

After PCR amplification, a 1% agarose gel electrophoresis was performed to verify the size and quality of the amplified fragments. As shown in Figure 3, distinct bands were observed in the expected lanes, with the backbone fragment appearing at approximately 1.5 kb and the color fragments (Light Pink, Blue, and Purple) resolving at approximately 500–800 bp. All bands were sharp and well-defined with no visible smearing or non-specific secondary bands, indicating high-specificity amplification with minimal off-target products. The absence of bands in the negative control lane further confirms that there was no contamination during the PCR setup. The fragment sizes observed are consistent with the expected sizes based on the primer design and template mUAV plasmid, confirming that the correct regions were successfully amplified. These results demonstrate that both the backbone and color fragment PCR reactions performed as expected, and that the purified products were of sufficient quality to proceed to Gibson Assembly.

3. DNA Purification and Quantification

1. Pipette 100 µL of DNA Binding Buffer into a centrifuge tube
2. Add 20 µL of PCR product
3. Mix briefly by vortexing
4. Transfer 120 µL of the mixture into separate columns with a collection tube
5. Centrifuge for 1 minute
6. Discard the flowthrough
7. Add 200 µL of DNA wash buffer to the column
8. Centrifuge for 1 minute
9. Repeat the last two steps
10. Transfer the column to a new tube
11. Discard flowthrough
12. Add 6 µL of nuclease-free water to the column matrix
13. Allow to sit for 2 minutes
14. Centrifuge for 1 minute
15. Store and save

Part Two

Materials (Credit to Lousia)

Items used:

• P1000 pipette with 1000 µL tips
• P20 pipette with 10 µL tips
• PCR Tubes

Biological materials:

• Purified Fragments
• Gibson Assembly Master Mix
• Nuclease-Free Water
• LB-Agar plates with Chloramphenicol
• SOC Growth Medium
• DH5α competent cells

Machines used:

• Thermal Cycler
• Shaking Incubator
• Waterbath set to 42°C

Part 1: Setting Up Gibson Assembly

We set up reactions in the proportions shown below for each color fragment, then incubated at 50°C for 30 minutes in a heat block, followed by adding 100 µL of nuclease-free water to dilute each sample.

Part 2: Transformation

1. Transfer 20 µL of competent cells to each tube
2. Transfer purified assembly products into each tube (8 total: 3 Light Pink, 3 Blue, 3 Purple)
3. Incubate on ice for 30 minutes

Figure 4. Tubes incubating on ice during the transformation step. Each tube is labeled by color and sample number.

4. Heat shock the cells at 42°C for 45 seconds immediately after the ice bath
5. Add 100 µL of SOC media to each tube
6. Allow growth in a shaking incubator for 1 hour
7. Transfer 100 µL from each tube to the appropriate plate and spread using plating beads or a plastic spreader
8. Incubate plates at 37°C for 72 hours

Part 3. Results

Figure 5 (A-H). LB-Agar plates with Chloramphenicol selection showing colony growth after transformation. Plates were labeled by color fragment condition Blue (B), Light Pink (LP), and Purple (Pu) at varying dilutions (Subject to correction with further observation). Interestingly, all colonies grew out purple-blue regardless of which color fragment was used. This may be because the insert DNA was not incorporated at the right ratio relative to the backbone, causing cells to express the backbone’s default color instead.

Week 7 Lab: Neuromorphic Circuits

Overview

In this lab, we designed two neuromorphic genetic circuits using the HTGAA 2026 Genetic Circuit Design Template and simulated their behavior using the Biocompiler-Predict tool. Both circuits are built from endoribonuclease-based sequestrons — the fundamental building blocks of intracellular artificial neural networks (IANNs) — and are intended for transfection into HEK293 cells via Lipofectamine 3000 and execution by an OT-2 liquid handling robot.

Key components used:

• Csy4 — a CRISPR endoribonuclease that cleaves mRNA at its recognition sequence
• CasE (EcoCas6e) — a second orthogonal endoribonuclease for independent mRNA cleavage
• PgU — a constitutive expression construct
• mNeonGreen, mKO2, eBFP2 — fluorescent protein reporters (green, orange, blue)
• _rec_ notation indicates a recognition site (e.g., Csy4_rec_mNeonGreen = mNeonGreen mRNA with a Csy4 cleavage site)

Circuit 1: “MyCircuit” (L-shape response)

Design rationale

This circuit implements a single-layer perceptron where two inputs (X₁ and X₂) each produce an endoribonuclease that negatively regulates a shared fluorescent output. The goal was to achieve an L-shaped dose–response surface: the output (mNeonGreen) should be high only when both inputs are low.

Analogy: Think of it like two faucets draining a bathtub. If either faucet is open (high X₁ or high X₂), water drains out and the tub level drops. The tub is full only when both faucets are closed.

Circuit design table

Total DNA: 700 ng

How it works

1. X₁ input delivers Csy4 endoribonuclease DNA (150 ng) along with mKO2 (orange fluorescent protein, 100 ng) as a transfection marker to verify X₁ delivery.
2. X₂ input delivers CasE endoribonuclease DNA (150 ng) along with eBFP2 (blue fluorescent protein, 100 ng) as a transfection marker for X₂.
3. Output layer consists of mNeonGreen mRNA with recognition sites for both Csy4 (Csy4_rec_mNeonGreen, 100 ng) and CasE (CasE_rec_mNeonGreen, 100 ng). Both endoribonucleases independently cleave the output mRNA.

When X₁ is high → more Csy4 is produced → more mNeonGreen mRNA is cleaved → output decreases. When X₂ is high → more CasE is produced → more mNeonGreen mRNA is cleaved → output decreases. When both are low → minimal cleavage → mNeonGreen output is maximal.

Predicted behavior

Figure 1: Biocompiler-Predict simulation of MyCircuit. The heatmap shows the predicted mNeonGreen output (Prediction Value) as a function of X₁ and X₂ concentrations. High output (dark blue, ~0.65–0.70) is concentrated along the left edge where X₁ is low. The L-shaped pattern confirms that the circuit acts as an approximate NOR-like function: output is highest when inputs are minimal.

Interpretation

The simulation reveals that X₁ (Csy4) has a stronger suppressive effect on the output than X₂ (CasE), as evidenced by the sharp drop-off along the X₁ axis compared to a more gradual decline along X₂. This asymmetry likely reflects differences in the catalytic efficiency and binding affinity of Csy4 versus CasE for their respective recognition sequences on the mNeonGreen mRNA. The L-shaped pattern is consistent with a weighted NOR gate where the X₁ weight is larger than the X₂ weight.

Circuit 2: “RF” (Rectified function)

Design rationale

This circuit implements a more complex multilayer architecture with cross-regulation between endoribonucleases. The goal was to achieve a rectified function — an output that increases monotonically with X₁ while remaining relatively insensitive to X₂, similar to a ReLU (rectified linear unit) activation function in machine learning.

Analogy: Imagine a volume knob (X₁) that smoothly turns up the music, while a second knob (X₂) has little effect because its signal gets cancelled out by internal feedback. The circuit “learns” to listen to one input and ignore the other.

Circuit design table

Total DNA: 700 ng

How it works

This is a multilayer circuit with cross-inhibition between the two endoribonucleases:

1. X₁ input delivers CasE (100 ng) and a reporter CasE_rec_Csy4_rec_mKO2 (50 ng) — an mKO2 mRNA that can be cleaved by both CasE and Csy4, acting as a dual-regulated node.
2. X₂ input delivers Csy4 (100 ng) and Csy4_rec_mNeonGreen (50 ng) — mNeonGreen output that is negatively regulated by Csy4.
3. Bias layer creates a rich cross-regulatory network:
   • PgU (100 ng) — constitutive expression baseline
   • CasE_rec_Csy4 (75 ng) — Csy4 mRNA with a CasE recognition site (CasE cleaves Csy4 mRNA)
   • Csy4_rec_CasE (75 ng) — CasE mRNA with a Csy4 recognition site (Csy4 cleaves CasE mRNA)
   • PgU_rec_CasE (75 ng) — constitutive mRNA regulated by CasE
   • PgU_rec_Csy4 (75 ng) — constitutive mRNA regulated by Csy4

The cross-inhibition (CasE_rec_Csy4 and Csy4_rec_CasE) creates a mutual negative feedback loop between the two endoribonucleases. This effectively implements a winner-take-all competition: when X₁ drives CasE production, CasE degrades Csy4 mRNA, further reducing Csy4 levels and amplifying the X₁ signal. The result is a rectified response that primarily follows X₁.

Predicted behavior

Figure 2: Biocompiler-Predict simulation of the RF circuit. The heatmap shows a smooth left-to-right gradient where output increases monotonically with X₁ (left axis = low, right axis = high). The output ranges from ~0.05 (white, low X₁) to ~0.55 (dark blue, high X₁). The response is largely independent of X₂, confirming the rectified function behavior.

Interpretation

The RF circuit successfully achieves a unidirectional dose–response: output scales with X₁ concentration while remaining approximately flat across X₂ values. This behavior arises from the mutual antagonism between Csy4 and CasE in the bias layer. When X₁ increases CasE levels, CasE degrades the Csy4_rec_CasE mRNA (reducing Csy4 production), which in turn reduces degradation of CasE mRNA — a positive feedback amplification of the X₁ signal. Meanwhile, X₂-driven Csy4 is counteracted by CasE from both the X₁ input and the bias layer, preventing X₂ from significantly influencing the output.

The smooth gradient (rather than a sharp threshold) reflects the analog nature of the IANN — the circuit computes a continuous function rather than a binary switch.

Comparison of the two circuits

Methods

Circuit design (Day 1)

1. Circuits were designed using the HTGAA 2026 Genetic Circuit Design Template (Google Sheet).
2. Part names followed the conventions in the HTGAA 2026 Genetic Circuit Part Names list.
3. All concentrations were set to 50 ng/µL (with one exception: Csy4 in MyCircuit at 40 ng/µL).
4. Circuit behavior was simulated using the Biocompiler-Predict tool, which generates heatmaps of predicted output across the X₁–X₂ input space.
5. Completed spreadsheets were uploaded via the Google Form submission.

Transfection and imaging (Day 2)

1. HEK293 cells were transfected using Lipofectamine 3000 with the designed plasmid mixes.
2. An OT-2 liquid handling robot in the Weiss Lab (NE-47, MIT campus) executed the transfection protocol based on our uploaded spreadsheet.
3. Fluorescence readout of mNeonGreen, mKO2, and eBFP2 will be measured after 24–48 hours of incubation.

Key takeaways

• Analog beats digital: Both circuits produce continuous, graded outputs rather than binary on/off responses — demonstrating the fundamental advantage of IANNs over traditional Boolean genetic circuits.
• Weight tuning via DNA dosage: The behavior of each circuit was tuned entirely by adjusting the nanogram amounts of each plasmid. No new genetic parts were needed — only different ratios of the same library components.
• Cross-inhibition enables complex functions: The RF circuit shows that mutual antagonism between endoribonucleases can create winner-take-all dynamics, allowing one input to dominate. This is a biological implementation of competitive inhibition analogous to lateral inhibition in neural circuits.
• Simulation before wet lab: The Biocompiler-Predict tool allowed us to iterate on circuit designs computationally before committing to expensive and time-consuming wet lab experiments.