Week 1 HW: Principles and Practices

Biological Engineering Application: Flavor-Engineered Yeast for Gastronomic Innovation

I want to develop genetically engineered yeast strains that produce novel flavor compounds during fermentation, creating breads, cheeses, wines, and fermented foods with entirely new sensory profiles.

Why this matters: While traditional breeding and natural yeast selection have given us incredible diversity—from sourdough’s tang to wine’s terroir—we’re limited by what wild Saccharomyces and other fungi naturally metabolize. Imagine baker’s yeast that produces vanilla notes without vanilla beans (addressing supply chain fragility), or cheese cultures that synthesize complex umami compounds usually requiring months of aging. This could democratize access to “luxury” flavors while reducing resource-intensive traditional production methods.

This connects to my background in environmental science and bioinformatics: using directed evolution and metabolic engineering, we could design strains that not only create desirable flavors but potentially upcycle waste streams (agricultural byproducts, water recycling plant outputs) as fermentation feedstocks—aligning with my interests in biomanufacturing and circular economies.

Governance/Policy Goals & Sub-Goals

Drawing from the synthetic genomics framework but adapting it for environmental release of consumer biotechnology, my primary goals are:

1. Environmental Protection & Biodiversity Conservation

• Sub-goal A: Prevent engineered yeast from establishing in wild ecosystems where they could outcompete native microbiota or transfer engineered genes horizontally
• Sub-goal B: Ensure rapid detection and response if environmental release occurs (containment failure, improper disposal of spent yeast)

2. Lab & Manufacturing Safety

• Sub-goal A: Prevent occupational exposure to novel metabolic byproducts that could be allergenic or toxic at scale
• Sub-goal B: Secure supply chains against intentional misuse (though flavor engineering seems low dual-use risk, the platform technologies—promoters, CRISPR tools, directed evolution systems—could be repurposed)

3. Equitable Access & Constructive Application

• Sub-goal A: Prevent patent lock-in that would concentrate flavor innovation in large corporations
• Sub-goal B: Ensure small artisan producers can access safe, validated strains without prohibitive regulatory burdens

Three Governance Actions

Action 1: Mandatory “Biocontainment-By-Design” Standards for Environmental Release

Actor: FDA (food safety oversight) + EPA (environmental release) collaboration

Purpose: Current GRAS (Generally Recognized as Safe) approval for food microbes doesn’t require genetic stability or kill-switch mechanisms. I propose requiring engineered yeasts intended for any release (including commercial food production where disposal occurs) to incorporate synthetic auxotrophies—strains that require supplemented nutrients unavailable in nature to survive.

Design:

• Technical: Engineered dependence on synthetic amino acids (inspired by the “Lysine Contingency” concept from Jurassic Park, but biologically real—see Church question below)
• Regulatory: FDA/EPA joint guidance document requiring demonstration of auxotrophy stability over >100 generations
• Implementation: Phase-in over 5 years; small artisan producers (<1000L/year) get delayed compliance but must use certified waste treatment (autoclaving) for spent yeast

Assumptions I might have wrong:

• That synthetic auxotrophies are genetically stable (evolutionary pressure might select for reversion)
• That kill switches don’t affect fermentation performance or flavor profiles
• That environmental establishment risk is high for S. cerevisiae (it might actually be low in many niches, but horizontal gene transfer to wild yeasts remains concerning)

Risks of Failure & “Success”:

• Failure: Reversion mutants establish in environment; gene flow to wild yeasts creates “super” competitors
• Success unintended consequences: If containment works too well, it becomes a barrier to open innovation—only well-funded labs can engineer stable auxotrophies, centralizing power; or, if we require extreme containment for all GMOs, we stifle beneficial environmental applications (bioremediation yeasts)

Action 2: Open-Source “Safe Strain” Repository with Pre-Cleared Regulatory Pathways

Actor: Non-profit consortium (modeled after Addgene or the iGEM Registry) + USDA/FDA fast-track approval

Purpose: Address the equity goal—currently, regulatory approval for novel GMOs costs $1-10M, accessible only to corporations. Create a public-domain collection of “chassis” yeast strains with pre-installed safety features and cleared regulatory status that any producer can modify.

Design:

• Technical: Curated set of ~20 GRAS yeast strains with multiple redundant auxotrophies, standardized modular cloning sites, and complete genomic sequences
• Governance: “Copyleft” licensing requiring derivative strains to remain open-source and safety-tested
• Funding: Initial NSF/NIH grant + membership fees from commercial users; academic users free
• Regulatory: FDA “master file” system where repository strains have pre-reviewed safety dossiers, reducing individual approval costs by ~90%

Assumptions I might have wrong:

• That open-source biology won’t create liability nightmares (who’s responsible if a modified strain causes harm?)
• That pre-cleared strains won’t stifle innovation (researchers might avoid novel chassis to save regulatory costs)
• That the “commons” approach prevents corporate capture (companies might still patent specific flavor pathways built on open chassis)

Risks of Failure & “Success”:

• Failure: Repository strains have hidden vulnerabilities; a widely-used chassis becomes an attractive bioterror target (monoculture risk); legal disputes over liability paralyze the system
• Success unintended consequences: If too successful, it creates a “two-tier” system—elite labs use cutting-edge proprietary strains while everyone else uses outdated open ones; or, rapid democratization leads to “biohacking” incidents before safety culture matures

Action 3: Mandatory Environmental Monitoring for Commercial Fermentation Facilities

Actor: State environmental agencies + municipal water treatment facilities

Purpose: Detect escape of engineered yeasts into wastewater and surrounding ecosystems before establishment occurs. Current food production regulations focus on human safety, not environmental surveillance.

Design:

• Technical: qPCR screening of wastewater for engineered signatures (synthetic promoters, auxotrophy markers) at facilities using GMO yeasts above threshold volumes
• Implementation: Quarterly testing required for facilities >10,000L/year; results reported to state EPA equivalents
• Funding: Industry pays testing costs; small producers exempted but encouraged via insurance discounts
• Response protocol: If detected, immediate source tracking, temporary production halt, and remediation (chlorination of wastewater)

Assumptions I might have wrong:

• That detection sensitivity is sufficient (yeast in complex wastewater matrices is hard to quantify)
• That environmental establishment is a gradual process we can catch early (it might be rare but catastrophic when it happens)
• That industry will comply rather than lobby against “unnecessary” costs

Risks of Failure & “Success”:

• Failure: False positives shut down innocent producers; false negatives miss real escapes; regulatory capture weakens standards
• Success unintended consequences: Creates surveillance infrastructure that could be repurposed for other monitoring (employee DNA, competitor espionage); or, perfect detection makes containment lax—“we’ll catch it if it escapes” mentality

Scoring Governance Actions Against Policy Goals

Scoring: 1=best, 3=worst, n/a=not applicable

Prioritized Recommendation

I recommend combining Action 1 (Biocontainment-by-Design) and Action 2 (Open-Source Repository), with Action 3 as a future phase-in once the infrastructure matures.

Rationale: Action 1 provides the fundamental safety architecture—without reliable biocontainment, we shouldn’t release engineered yeasts at all, regardless of monitoring. Action 2 addresses the equity and innovation goals that Action 1 alone would compromise (if only wealthy companies can afford containment engineering). Together, they create “safe by default, accessible to all” infrastructure.

Trade-offs considered:

• Cost vs. safety: Action 1 is expensive; Action 2 subsidizes access. The combination spreads costs across a community rather than individual producers.
• Prevention vs. response: I deprioritized Action 3 because detection without prevention feels ethically insufficient for environmental release—we shouldn’t “test the water” to see if we broke the biosphere. However, once Actions 1-2 are established, Action 3 provides valuable validation that containment works.

Audience: This recommendation is directed to Dr. Tara O’Toole, Under Secretary for Science and Technology at DHS, and Dr. Renee Wegrzyn, Director of ARPA-H, as they oversee biosecurity innovation and health security research programs that could fund the open-source repository and establish containment standards for consumer biotechnology.

Key uncertainties: We don’t know the actual environmental establishment risk of domesticated S. cerevisiae (it may be lower than feared, or gene flow to wild yeasts may be the real issue). We also don’t know if synthetic auxotrophies impose metabolic burdens that make commercial strains non-viable. Pilot studies in contained environments (wastewater treatment plants, industrial baking facilities) should precede broad policy.

Reflection on Week 1 Ethical Concerns

The class’s “cell as silicon switch” analogy is intellectually compelling but ethically unsettling. I found myself excited by the engineering logic—modularity, predictability, abstraction layers—yet troubled by what this framing obscures. Living systems evolve, have ecological contexts, and exist in webs of relationships that “switches” don’t.

The yeast flavor project crystallized this tension: I can engineer a metabolic pathway as cleanly as a circuit, but I cannot engineer the evolutionary stability of that pathway, nor predict its ecological interactions if released. The “switch” metaphor suggests control we don’t actually possess. This feels like a new ethical concern for me—epistemic humility in the face of biological complexity. We need governance that acknowledges uncertainty rather than pretending engineering precision eliminates it.

Governance action to address this: Mandatory “Evolutionary Impact Statements” for environmental release applications, requiring developers to model (and monitor) not just intended function but evolutionary trajectories and ecological interactions over multi-year timescales. This would force the engineering mindset to confront biological reality.

Homework Questions — Professor Jacobson

Nature’s machinery for copying DNA is called polymerase. What is the error rate of polymerase? How does this compare to the length of the human genome. How does biology deal with that discrepancy?

DNA polymerases have error rates around 10⁻⁹ to 10⁻¹⁰ per base pair (after proofreading by 3’→5’ exonuclease activity). The human genome is ~3×10⁹ base pairs.

Without correction, this would yield ~3-30 mutations per replication—catastrophic over organismal lifetimes. Biology deals with this through:

1. Proofreading (intrinsic to polymerase—catches ~99% of errors immediately)
2. Mismatch repair (post-replication surveillance by MSH/MLH proteins)
3. Redundancy (diploidy allows masking of recessive deleterious mutations)
4. Selection (organisms with too many errors don’t survive/reproduce)

The combined fidelity is ~10⁻¹¹, yielding ~0.03 mutations per replication—manageable evolutionary noise.

How many different ways are there to code (DNA nucleotide code) for an average human protein? How does this compare to the length of the human genome. How does biology deal with that discrepancy?

The genetic code is degenerate: 61 sense codons specify 20 amino acids. For a protein of length n, there are potentially 20ⁿ amino acid sequences, but 61ⁿ possible DNA sequences coding for a specific amino acid sequence (accounting for synonymous codons).

For an average human protein (~375 aa), there are ~3⁷⁵⁰ ≈ 10³⁵⁷ possible DNA sequences coding for the same protein—an astronomically large “sequence space.”

Why all these codes don’t work equally:

1. Codon usage bias: Different organisms prefer different synonymous codons (match tRNA abundances)
2. mRNA secondary structure: Some sequences form inhibitory hairpins
3. Cryptic regulatory elements: Some “silent” mutations create/destroy splice sites, promoters, or regulatory motifs
4. Translation kinetics: Rare codons cause ribosomal pausing, affecting co-translational folding
5. Chaperone interactions: Translation speed affects protein folding fidelity

Thus, “DNA → protein” is not a simple coding problem but an optimization across multiple constraints.

Homework Questions — Dr. LeProust

What’s the most commonly used method for oligo synthesis currently?

Phosphoramidite solid-phase synthesis, using automated synthesizers. This chemistry cycles through: deprotection → coupling → capping → oxidation, adding one nucleotide per cycle. It’s robust, parallelizable, and achieves >99.5% stepwise efficiency.

Why is it difficult to make oligos longer than 200nt via direct synthesis?

Cumulative error from imperfect stepwise efficiency. At 99.5% coupling efficiency per cycle, after n cycles the yield is (0.995)ⁿ. For 200nt: (0.995)²⁰⁰ ≈ 37% full-length product. Beyond 200nt, truncated failure sequences dominate, requiring difficult purification. Also, depurination during acidic deprotection steps increases error rates in long syntheses.

Why can’t you make a 2000bp gene via direct oligo synthesis?

Same problem magnified: (0.995)²⁰⁰⁰ ≈ 0.004% yield—essentially zero full-length product. Instead, we use assembly strategies:

• PCA (Polymerase Cycling Assembly): Overlapping ~60nt oligos prime each other, polymerase fills gaps
• Gibson Assembly: Exonuclease creates overhangs, annealing and ligation join fragments
• Golden Gate/Type IIS restriction enzymes: Directional scarless assembly

These “divide and conquer” approaches bypass the error accumulation of direct synthesis.

Homework Question — Professor Church

[Selected: Question 1 — Essential amino acids and the Lysine Contingency]

What are the 10 essential amino acids in all animals and how does this affect your view of the “Lysine Contingency”?

The 10 essential amino acids that animals cannot synthesize (must obtain from diet) are: arginine, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan, and valine (some sources vary slightly by species/age; arginine is conditionally essential).

Re-evaluating the “Lysine Contingency”:

In Jurassic Park, the “Lysine Contingency” was a fictional biocontrol—dinosaurs engineered to require supplemental lysine, ensuring they couldn’t survive if they escaped. This seemed clever but biologically naive: lysine is abundant in nature (legumes, meat), and engineered auxotrophies can revert.

However, the concept is biologically grounded. Real applications include:

• “Synthetic auxotrophy” in GMOs: Engineering dependence on synthetic amino acids not found in nature (e.g., p-aminophenylalanine, or “non-canonical” amino acids)
• Xenobiology: Expanding the genetic code with unnatural base pairs or amino acids creates orthogonal biological systems

My view: The Lysine Contingency is inspiring but insufficient alone. Single-auxotrophy fails because (1) metabolic bypasses evolve, (2) the “essential” nutrient may not be truly scarce, and (3) horizontal gene transfer can restore synthesis. For robust biocontainment, we need redundant, multi-layered systems—multiple synthetic auxotrophies, genetic recoding (changing stop codons to require unnatural tRNAs), and physical containment. This aligns with my Action 1 governance proposal: containment must be “by design” and multi-layered, not relying on single points of failure.

Submitted by Amaya Sarmiento, HTGAA 2026