Homework #1 — Biomanufacturing Proposal

(1) What I want to biomanufacture, and why

I want to biomanufacture a living diagnostic: an engineered strain of Escherichia coli Nissle 1917 (a well-characterized, already-human-approved probiotic chassis) that senses gut inflammation in situ and reports it through a recoverable signal in stool — ÌṢỌ Sentinel EcN.

The specific molecule being sensed is tetrathionate, an oxidized sulfur species that the gut does not normally contain in meaningful quantities but that is produced when host-derived reactive oxygen species (ROS), generated during inflammation, oxidize luminal thiosulfate. This chemistry was established by Winter et al. (2010) in a Salmonella infection model and has since been shown by Daeffler et al. (2017) and Riglar et al. (2017) to correlate with inflammation more broadly, including in IBD mouse models, using engineered TtrS/TtrR two-component sensing circuits built into E. coli.

My motivation is both personal and professional. As someone completing a medical internship in Nigeria and building a research career around understanding why individuals respond differently to molecular interventions, I'm interested in non-invasive, low-cost diagnostic tools that don't depend on endoscopy, colonoscopy, or lab infrastructure that's scarce in many of the settings I've trained in. IBD and related gut inflammatory conditions are difficult and expensive to monitor longitudinally — colonoscopy is invasive, costly, and impractical for frequent repeat testing, which is exactly what's needed to track disease flares or treatment response over time. A swallowable bacterial sensor that reports through stool is cheap to produce, requires no specialized clinical procedure to deploy, and could in principle be read out with something as simple as a colorimetric strip or fluorescence reader rather than a hospital lab.

This product is for patients managing IBD and clinicians monitoring them, and for researchers studying gut inflammation dynamics who currently rely on invasive, low-frequency sampling.

I'm deliberately scoping this as a diagnostic/reporting tool, not a therapeutic delivery system. The sensing circuit (TtrS/TtrR responding to tetrathionate) is well-validated in the literature; a therapeutic payload-release function would require a second layer of biology (release mechanism, payload stability, dosing control) that I have no validation data for and that would overclaim what this proposal can actually support.

(2) What living system would I use?

Chassis: E. coli Nissle 1917 (EcN), a probiotic strain already used clinically and considered one of the best-characterized safe gut-colonizing chassis for engineered bacterial therapeutics/diagnostics (this is the same chassis Riglar et al. used for long-term gut monitoring in mice).

How it works, at a high level:

• Input: tetrathionate present in the gut lumen during inflammation (produced from host ROS oxidizing thiosulfate — I am not engineering the production of tetrathionate, only sensing it).
• Biological process: a two-component signaling system — TtrS (membrane-bound sensor histidine kinase) detects extracellular tetrathionate and autophosphorylates; it transfers the phosphate to TtrR (the cognate response regulator), which then activates a tetrathionate-responsive promoter (PttrB).
• Output: the activated promoter drives expression of a reporter gene. Because this needs to be recoverable from stool rather than read in real time inside the body, I am assuming a reporter with a stable, easily extractable readout — e.g., a chromoprotein (such as a Betacoral/blue or similar pigment) rather than a fluorescent protein, since pigments can be visually or spectrophotometrically read from a stool sample without a fluorescence reader, which better matches a low-resource diagnostic use case than sfGFP would.
• Containment: an auxotrophic safety mechanism (ΔdapA, requiring exogenous diaminopimelic acid for cell wall synthesis) ensures the strain cannot persist or proliferate outside a controlled dosing context, addressing the obvious biocontainment concern of releasing engineered bacteria into a patient's gut and into wastewater afterward.

Assumptions I'm making explicitly: I'm assuming the TtrS/TtrR circuit, when placed in EcN rather than the lab E. coli strains used in the cited literature, retains meaningful sensitivity and specificity to tetrathionate — this is a real, flagged risk (see Part 5), since the original iGEM team's implementation of this same circuit reported weak functional performance in their hands. I'm also assuming a chromoprotein reporter is swappable in place of the sfGFP reporter used in prior validated circuits without disrupting promoter activation kinetics — a reasonable assumption since the reporter sits downstream of the same regulatory logic, but not something I've validated.

(3) How would I use automation?

Given that I am currently a remote HTGAA participant without walk-up access to a wet lab or a Cloud Lab contract, I'm proposing desktop liquid-handling automation (e.g., Opentrons OT-2) as the realistic automation layer for this project, rather than a large-scale Cloud Lab platform I would have no practical way to operate or justify the cost of as an individual remote participant.

What I'd automate: strain construction and circuit-variant screening — specifically, automated assembly reaction setup (combining backbone + parts for Golden Gate or Gibson assembly), transformation plating, and colony screening across multiple TtrS/TtrR promoter-strength variants and reporter choices. This is the step in the workflow most worth automating because it's the most repetitive and error-prone part of the process by hand: testing several promoter variants and reporter constructs in parallel to find one that gives a strong, specific tetrathionate response (directly addressing the sensitivity/specificity risk above) requires many near-identical pipetting steps, which an Opentrons platform handles with far better reproducibility and far less hands-on time than manual bench work — important given I'd be running this as a side project alongside clinical duties.

I'm not proposing to automate large-scale fermentation/production at this stage, since that belongs to the "scale" question below, not the prototyping stage.

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

(a) How much?

I'm scoping this to a diagnostic validation cohort, not population-wide clinical deployment — this is the honest scale for a first-generation diagnostic tool that hasn't yet been clinically validated, and it avoids the trap of estimating manufacturing requirements for a claim I can't yet support.

Target: enough engineered EcN doses for a 100-person IBD biomarker validation study, with each participant dosed once daily for a 2-week monitoring window (a reasonable timeframe to capture at least one inflammatory flare/remission cycle, consistent with how IBD activity is typically tracked via biomarkers like fecal calprotectin in clinical practice).

• 100 participants × 14 days × 1 dose/day = 1,400 total doses
• Typical oral probiotic dosing is on the order of 10⁹–10¹⁰ CFU per dose (consistent with commercial EcN product dosing, e.g. Mutaflor).
• Assume 10¹⁰ CFU/dose → total CFU needed = 1,400 × 10¹⁰ = 1.4 × 10¹³ CFU

Converting to culture volume: E. coli in rich media (e.g., LB) at stationary phase typically reaches ~10⁹ CFU/mL.

• Volume needed = 1.4 × 10¹³ CFU ÷ 10⁹ CFU/mL = 1.4 × 10⁴ mL ≈ 14 liters of saturated culture

This is a modest, benchtop-to-small-bioreactor scale — well within reach of a single 10–20 L fermenter run, not an industrial manufacturing problem. That's appropriate for a validation-cohort target and is exactly why I'm not trying to size this for "everyone with IBD," which would require clinical trial data this proposal doesn't have.

(b) How long?

Prototyping stage (Opentrons-scale): circuit construction and variant screening doesn't need 14 L — a handful of 1–5 mL cultures per variant, screened in parallel (e.g., 24–96 variants in a 96-well format) is enough to validate circuit function and pick the best-performing construct. Using an Opentrons OT-2 for assembly + plating, plus standard overnight growth, a full build-test-learn cycle (assemble → transform → grow → screen) takes roughly 3–5 days per round, and I'd expect to need 2–3 rounds to converge on a working circuit (accounting for the known risk that this exact sensing circuit has previously underperformed in other teams' hands) — so roughly 2–4 weeks for prototype validation.

Scale-up stage (14 L production): once a validated construct is selected, scaling a single strain from a colony to a 14 L batch is a standard fermentation timeline: seed culture (overnight, ~16 h) → scale-up through intermediate volumes (1–2 days) → final fermentation run to stationary phase (~16–24 h) → harvest/formulation (encapsulation or lyophilization for oral dosing, another 1–2 days). Total: roughly 4–6 days per full-batch production run, which comfortably covers the 1,400-dose target in a single run with some margin, since most standard lab/pilot fermenters exceed 14 L capacity.

So: prototype validation (2–4 weeks) is actually the longer, rate-limiting phase, not production — which is a very different bottleneck profile than I'd expect for a product requiring industrial-scale output, and is consistent with this being correctly scoped as a diagnostic tool project rather than a bulk-manufacturing project.

(5) What don't I know yet?

1. Will the TtrS/TtrR circuit actually give a strong, specific signal in EcN, given it's been reported to perform poorly in at least one other team's implementation? The iGEM SHSBNU_China team's 2017 build of this same tetrathionate-sensing part reported weak functional output. I don't yet know whether that reflects a fundamental limitation of the circuit, a chassis-specific issue (their work wasn't in EcN), or an implementation problem specific to their build. How I'd find out: run the promoter-variant screening described above (Opentrons-automated), testing multiple PttrB promoter-strength variants and TtrS/TtrR expression levels against a dose-response curve of synthetic tetrathionate in vitro before ever testing in an animal/gut model, to characterize sensitivity and dynamic range empirically rather than assuming the published circuit works as-is.

2. Is tetrathionate specific enough to inflammation, or will it cross-react with other gut conditions/diet-driven sulfur metabolism, producing false positives? Published work establishes tetrathionate as inflammation-associated, but gut sulfur chemistry is diet-sensitive and the signal's specificity in a free-living human gut (versus a controlled mouse model) is unknown. How I'd find out: this requires an actual human/animal validation study — not something I can resolve at the circuit-design stage — but I could at minimum stress-test the in vitro circuit against a panel of other oxidized sulfur species (thiosulfate alone, sulfate) to characterize cross-reactivity before moving to any biological validation.

3. Will a chromoprotein reporter (needed for stool-readable output) preserve the same activation kinetics and detection sensitivity as the sfGFP reporter used in the original validated circuits? I'm assuming reporter-swapping is low-risk, but chromoproteins and fluorescent proteins differ in maturation time and expression-level sensitivity, and I have no direct evidence this swap preserves circuit performance. How I'd find out: build both reporter versions in parallel during the Opentrons screening round and directly compare their dose-response curves against the same tetrathionate standard, rather than assuming equivalence.