Week 1 HW: Final Project Ideas

Biomanufacturing an anti-troponin Fab fragment for point-of-care heart diagnostics

What would I like to biomanufacture and why?

Target product: a recombinant Fab (antibody fragment) that binds to cardiac troponin I (cTnI — a specialized protein found exclusively in heart muscle cells) to be used as the detection reagent in a point-of-care lateral flow immunoassay test strip for diagnosing myocardial infarctions. The end product would work the same as a pregnancy test — for sensitive, rapid, and user-friendly quantification.

Motivation: Troponin I is located inside the heart’s muscle cells, where it normally should be, but damage to the cells causes troponin to leak into the blood. This is why cTnI is invaluable for cardiac diagnosis and assessing heart attack severity. When I was first introduced to the study “AI-driven lateral flow immunoassay for point-of-care detection of cardiac biomarkers in acute myocardial infarction” (Lu et al., 2026), I quickly realized how important point-of-care technology is becoming these days, and how migrating from slower turnaround times to instant results could speed up the decision to treat. According to the study, the prototype showed quantitative results in about 8 minutes, a roughly 47% improvement in detection efficiency over conventional methods (Lu et al., 2026). The antibody reagent itself is usually the expensive, hard-to-scale component of these devices, which is exactly the piece a biomanufacturing perspective can address. As a biomedical engineer living in a country where underdeveloped areas are common (Honduras), I firmly believe in the phrase “time is muscle,” and in how reliable and low-cost technologies can save countless lives.

Target audience: rural and low-resource clinics, and emergency department triage — basically any setting where lab-based troponin testing is too slow or unavailable.

What living system would I use?

Chassis: E. coli strain BL21(DE3), which is widely used for the expression of recombinant proteins in biomanufacturing. This bacterium would be engineered to express and secrete the anti-cTnI Fab fragment into the periplasm of the cell (its oxidizing environment promotes the bond between troponin I and the anti-cTnI fragment).

Why a microbial cell? This project proposes a diagnostic reagent, where the sole focus is high-affinity binding to ensure its functioning. Fab fragments lack the Fc region and can fold correctly without glycosylation, so bacterial expression is sufficient. This is a well-established approach for diagnostic-grade antibody fragments and is far cheaper and faster to scale than mammalian cell culture.

1. In: glucose/rich growth medium for high-density growth, an inducer (IPTG) to turn on protein production and make it constant, pH, and controlled temperature.
2. Process: The specialized bacterial strain is grown to high density through fed-batch fermentation, then expression is induced after glucose depletion. Heavy and light chains are co-expressed with a periplasmic signal peptide so the Fab folds correctly in the oxidizing periplasmic environment needed for its disulfide bonds.
3. Out: harvested cells/media → periplasmic extraction or cell lysis → affinity purification using Protein L → purified Fab ready for conjugation and application onto a lateral flow membrane. Functionalization follows citrate-reduction synthesis of gold nanoparticles, surface functionalization, and covalent antibody conjugation via EDC/NHS coupling chemistry, as stated in the study for the lateral flow test strip.

Assumptions: the anti-cTnI Fab fragment is completely compatible with bacterial expression, and periplasmic secretion gives an acceptable fraction of correctly folded, active protein.

To consider: a cell-free protein synthesis (CFPS) system using E. coli lysate/extract directly in a reaction, rather than growing live cells, as an alternative chassis. Addition of whey permeate to E. coli-based cell-free extracts has been reported to enhance cell-free protein synthesis yield by about 50% (Rasor et al., 2021).

Automation

Technology: a desktop liquid-handling robot (Opentrons OT-2) paired with a plate reader to corroborate binding, applied to the antibody screening and expression stage of the workflow.

What I would automate: the first step, and one where automation would be ideal, is to screen anti-cTnI Fab fragments and growth conditions — also to ensure expression conditions have been met to find the combination that gives the best yield and the best-folded, highest-affinity antibody. These steps include colony picking, small-culture growth, induction, lysis, and ELISA binding assays, which a liquid-handling robot is perfect for. At the manufacturing stage, automation would shift toward downstream processing with an automated FPLC system for purification runs.

For the cell-free alternative, a programmable digital microfluidics platform (Nuclera) would be the best fit, since cell-free reactions don’t require cell growth or an induction step, so each test could be run and screened on-chip directly.

How much would I need to make, and how long would it take?

Scale definition: enough Fab reagent to supply point-of-care troponin strips for every emergency department chest-pain evaluation in Massachusetts in one year. Nationally, chest pain is the second most common reason for an ED visit in the U.S. — 6.5 million visits per year out of a U.S. population of roughly 335 million. Massachusetts has a population of about 7.15 million (2025). Applying the national rate to Massachusetts gives an estimated demand:

National chest-pain-visit rate:
6,500,000 ÷ 335,000,000 ≈ 1.94% of population/year

Applied to Massachusetts:
7,150,000 x 1.94% ≈ 138,800 strips/year

For the antibody loading per strip, published lateral flow assay development protocols typically formulate the solution at about 1 mg/mL, with a separate amount used to functionalize the test strip. Based on these figures, I estimate roughly 1 µg of total Fab (capture + detection) per strip.

138,800 strips/year x 1 µg Fab/strip
= 138,800 µg ≈ 0.139 g (139 mg) of Fab fragment per year

How long? Fed-batch E. coli periplasmic Fab expression has been demonstrated at titers around 100 mg/L of culture in bench-scale (10 L) and pilot-scale (300 L) fermenters.

Volume of culture needed:
139 mg ÷ 100 mg/L ≈ 1.4 L of culture

This is designed to fit in a 2- to 10-L benchtop bioreactor; no parallel reactors needed. At ~3 days/batch (growth + induction + harvest + purification), Massachusetts’s full annual demand could be produced in a single ~3-day run.

Honduras

Honduras has the highest ischemic heart disease (IHD) and cardiovascular mortality rate in Central America, with an average death rate of 82.7 per 100,000 people, yet Honduran Ministry of Health (SESAL) figures report just over 100 heart attack deaths per year nationwide. This gap aligns with a well-known problem in resource-limited settings: underdiagnosis of acute MI, often driven by the same access and lab-turnaround barriers this project targets.

Suspected chest-pain evaluations/year:
1,300,000 x 1.94% ≈ 25,200 strips/year

Fab needed: 25,200 x 1 µg/strip ≈ 0.025 g (25 mg)/year

Culture volume: 0.025 g ÷ 100 mg/L ≈ 0.25 L (250 mL)
→ producible in a single small shake-flask or benchtop
  fermenter run, well under 1 batch, ~2-3 days total

What don’t I know yet?

• Antibody required per strip: my ~1 µg/strip figure is inferred from published lateral flow formulation concentrations, not a number specific to a troponin assay. I’d need to prototype actual strips and titrate the true capture/detection antibody load required for a clinically useful sensitivity cutoff.
• Correctly folded, active fraction from bacterial expression: periplasmic Fab expression can partition unpredictably between periplasm, culture medium, and insoluble inclusion bodies. I’d need to experimentally measure what fraction of expressed protein is soluble and competent for antigen binding, since this could substantially change the effective fermentation volume needed.
• Scale-up fidelity: fermentation titers measured in small pilot runs (10 L) don’t always hold at larger production scale. I’d need to run and validate titer at the actual production bioreactor scale before trusting the timeline.

References

Lu, Y., Peng, X., Yin, Z., Fan, X., Fan, J., Mi, Y., & Li, G. (2026). AI-driven lateral flow immunoassay for point-of-care detection of cardiac biomarkers in acute myocardial infarction. Mikrochimica Acta, 193(5), 323. https://doi.org/10.1007/s00604-026-08054-y

Cleveland Clinic. (2022, marzo 17). Troponin test. Cleveland Clinic. https://my.clevelandclinic.org/health/diagnostics/22770-troponin-test

Rasor, B., Vögeli, B., Landwehr, G., Bogart, J., Karim, A., & Jewett, M. (2021). Toward sustainable, cell-free biomanufacturing. Current Opinion in Biotechnology. https://doi.org/10.1016/j.copbio.2020.12.012

Final Project

Week 02: DNA design

HTBAA Homework #2 — AntiCTnI_Fab_ExpressionPlasmid

Part 0: Setup in Kernel

Repository: HTBAA26_WilliamandMary. Folder: Marcela Mejia.

Part 1: Expression Target

Molecule: A recombinant Fab (antibody fragment) directed against human cardiac troponin I (cTnI). A Fab is a protein made of two separate chains — a heavy chain fragment (VH-CH1) and a light chain (VL-CL) — that associate after folding to form one functional antigen-binding unit. A Fab is chosen over a full antibody because it is smaller, simpler to express in bacteria, and sufficient for the diagnostic/binding function needed.

Sequence and source: The construct is a chimeric Fab. The variable domains (VH and VL) come from PDB 8DZV, a chicken anti-cardiac Troponin I antibody–peptide complex structure (Islam et al., 2023). These are fused to human constant domains: CH1 from human IgG1, UniProt P01857 (IGHG1), for the heavy chain, and the human IGLC (lambda) constant region for the light chain. Each chain’s protein sequence was reverse-translated to DNA and codon-optimized for E. coli, then verified by back-translation to confirm an exact match to the source protein sequence before being built as Kernel CDS parts. I chose this chimeric design because chicken antibodies are commonly raised against highly conserved mammalian proteins like cardiac troponin. Human cTnI looks too similar to a mammal’s own troponin to trigger a strong immune response, so mammals make weak antibodies against it. Chickens are different enough from mammals that their immune system reacts strongly to human cTnI, producing good, high-affinity antibodies instead. So if we take the binding part of that chicken antibody (VH/VL) and attach it to a standard human antibody “body” (CH1/CL), the product has stronger binding.

Host/chassis: Escherichia coli, generic strain (no DE3/T7 RNA polymerase available), which ruled out T7-driven expression and drove the choice of a Lac promoter recognized by native E. coli RNA polymerase.

Product localization: Periplasm. Proper Fab assembly requires correct inter-chain disulfide bond formation, which is far more reliable in the oxidizing environment of the periplasm than in the reducing cytoplasm. This is why the design includes PelB (heavy chain) and OmpA (light chain) signal peptides, each fused in-frame to its respective chain to direct Sec-dependent periplasmic translocation.

Part 2: Genetic Parts

Both signal-peptide fusions were verified translation-frame-correct in Kernel’s sequence view: each chain’s cistron reads start-to-stop with no premature stop codons at the fusion junctions.

Part 3: Assembled Construct

AntiCTnI_Fab_ExpressionPlasmid is a periplasmic E. coli expression construct for an anti-cardiac Troponin I (cTnI) Fab fragment, driven by a single Lac promoter with the heavy chain (PelB-VH-CH1) and light chain (OmpA-VL-CL) translated independently from separate RBSs as a bicistronic operon. Both signal peptides direct their chains to the periplasm for proper folding and disulfide bond formation; the light chain carries a C-terminal 6xHis tag for Ni-affinity purification, and the construct is closed out with a BBa_B0015 terminator and a pUC-AmpR v3 backbone for high-copy propagation and ampicillin selection in E. coli.

5’→3’ order (as built): pUC-AmpR v3 backbone → Lac promoter → RBS1 → PelB → VH-CH1 → RBS2 → OmpA → VL-CL → 6xHis tag → BBa_B0015 terminator → back into the backbone.

Verification performed: The assembled sequence was checked base-by-base in Kernel’s translation view. Both signal-peptide junctions (PelB→VH-CH1 and OmpA→VL-CL) read as continuous, in-frame fusions with no stop codons at the junction, confirmed by the correct canonical leader-sequence amino acids (PelB: MKYLLPTAAAGLLLLAAQPAMA; OmpA: MKKTAIAIAVALAGFATVAQA) running straight into each chain’s variable-domain sequence. VH-CH1 terminates with a clean double stop codon (TGA, TAA) before RBS2, and VL-CL + 6xHis terminates with a clean double stop before the BBa_B0015 terminator. Total construct length: 3740 bp, consistent with the sum of all individual part lengths.

Part 4: Optimize and Quality-Check

Codon optimization. As a second, independent pass, VH-CH1 and VL-CL were each re-run through Kernel’s Coding Sequence Optimization tool. The protein sequence output was identical before and after (codon optimization is synonymous), but individual codons changed — for example GCG→GCA (Ala), GTG→GTT (Val), GGC→GGT (Gly), and TGC→TGT (Cys). Codon usage matters because E. coli’s tRNA pool is unevenly abundant across synonymous codons; a codon that is common in the human/chicken source sequences can be rare in E. coli, causing ribosomal stalling and reduced yield, while recoding to E. coli-preferred codons keeps translation efficient without altering the protein. No restriction-site avoidance or GC-content constraint was applied in this run.

Part 5: From Design to Build

How I would build it. At 3740 bp total, with a ~1.6 kb insert (promoter through terminator) cloned into a 2077 bp backbone, this construct is well within the size range for commercial whole-gene synthesis. I would order the insert as synthetic DNA from a vendor and join it to the linearized pUC-AmpR backbone using Gibson assembly. Gibson et al. (2009) describe a one-pot isothermal method that joins DNA fragments sharing short (~20–40 bp) overlapping ends using an exonuclease, polymerase, and ligase acting together, without restriction sites — well suited here since the construct combines several registry parts (promoter, RBS, signal peptides, terminator) with two custom CDS inserts and a reused backbone, and Gibson assembly avoids needing compatible restriction sites across all of them.

What synthesis constrains. Gene synthesis has non-trivial per-base error rates, and struggles with highly repetitive sequences, long homopolymer runs, and extreme GC content — none of which this construct has in excess, but the repeated use of the same RBS (BBa_B0034) twice and two independent signal-peptide/CDS fusions are worth flagging as regions to sequence-verify post-synthesis carefully. Carr et al. (2004) describe protein-mediated error correction (using mismatch-binding proteins like MutS to remove error-containing strands from a synthesis pool before amplification), which becomes especially relevant if any part of this construct were instead built from chip-synthesized oligo pools rather than vendor gene synthesis — chip-based methods (Tian et al. 2004; Kosuri et al. 2010; Kong et al. 2007) have substantially higher per-base error rates than column-based vendor synthesis and would need that kind of error correction step before assembly.

Prototype vs. scale. For an initial prototype — testing whether the Fab expresses and binds cTnI at all — ordering one synthesized insert and assembling a handful of clones by Gibson assembly is sufficient and fast. If this moved toward production scale, I would shift toward chip-based parallel oligo synthesis (Tian et al. 2004; Kosuri et al. 2010), which produces many DNA sequences in parallel far more cheaply per base than individual gene synthesis orders, paired with the error-correction step above to manage its higher error rate.

Part 6: What Don’t You Know Yet?

1. Will the chimeric chicken-VH/VL + human-constant-domain fusion fold and assemble correctly? This is a chimeric Fab, not a natural sequence, so the VH/CH1 and VL/CL domain interfaces have not been validated as compatible. How I’d find out: express the construct, extract the periplasmic fraction, and run non-reducing SDS-PAGE to check for a disulfide-linked heavy/light complex, followed by an ELISA against cTnI to confirm the assembled Fab actually binds antigen.

2. Will the two chains express in a balanced enough ratio for proper 1:1 Fab assembly? A single promoter driving two independently initiated cistrons (via RBS1 and RBS2) does not guarantee equal translation rates for the heavy and light chains. How I’d find out: run a Western blot post-induction distinguishing heavy and light chain bands (by size or epitope tag) to quantify relative expression levels, and if imbalanced, swap in a different-strength RBS on one cistron to rebalance.

3. Is codon optimization alone sufficient, or will mRNA secondary structure near the RBS/start codons hinder translation initiation? Two independent RBS-CDS pairs in one transcript raise the chance of unintended secondary structure forming between the two cistrons. Kernel’s RNA structure score tool returned N/A on this run, so this remains unverified in software. How I’d find out: test expression via an IPTG induction time-course and compare yield to a single-cistron control; if expression is lower than expected, use Kernel’s structure-scoring tool on the 5’ region around each RBS to check for the problem area.