Week 1 HW: Final Project Ideas

HTBAA Homework #1

Target: Cultivated bovine muscle meat grown in yeast-hydrolysate-supplemented medium, incorporating a synthetic gene network for stem-state control

1. What I want to biomanufacture, and why

I want to biomanufacture cultivated bovine muscle tissue, muscle mass grown from bovine satellite cells (the myogenic stem cells resident in skeletal muscle), intended as a cultivated meat product, using a growth medium that is partially reformulated with a protein hydrolysate derived from food-grade Saccharomyces cerevisiae (baker’s yeast), and incorporating a synthetic gene network that keeps the cells in a proliferative, undifferentiated state during expansion.

This idea draws directly on two pieces of my own work this year, as part of my MRes in Synthetic Biology at UCL. First, my MRes research project evaluated whether a yeast-derived hydrolysate (YH) could partially substitute for the amino acid component of Dulbecco’s Modified Eagle Medium (DMEM), the standard basal medium (BM) used in mammalian cell culture, in order to reduce the cost and animal-dependency of cultivated meat production. Second, in an earlier coursework essay for this same MRes, I designed, as a creative exercise rather than an experimentally built construct, a synthetic gene network (SGN), which I called SGN X, intended to keep bovine satellite cells in a stem-like, non-differentiated state under ex vivo culture conditions by sensing oxygen tension. Bringing these two together, a cheaper and less animal-dependent medium, together with a genetic circuit that actively maintains the cells’ proliferative capacity in that medium, is the biomanufacturing idea I want to propose here.

Image 1 - Brief biomanufacturing process diagram of cultivated meat production

The motivation is directly utilitarian. Cultivated meat is frequently proposed as a lower-footprint alternative to conventional livestock farming, but its commercial viability is bottlenecked by the cost and animal-derived nature of cell culture media, principally fetal bovine serum (FBS) and the amino acids supplied by DMEM. My own MRes research indicated that serum is by far the largest single cost driver of standard culture medium, with DMEM itself a further large, purely synthetic cost component. Yeast hydrolysate is attractive because it is cheap, scalable, food-grade, and not animal-derived, since S. cerevisiae fermentation is already an industrially mature process. Separately, a persistent challenge in myoblast expansion is that cells can drift toward premature differentiation under culture stress, which reduces the proliferative yield needed to reach meaningful biomass; this is what the genetic circuit component is meant to address.

This is for two overlapping audiences: the cultivated meat industry, which needs cheaper, more scalable, less animal-dependent media components and better control over cell fate during expansion, and more broadly, anyone interested in decoupling protein production from land-intensive animal agriculture.

2. What living system would I use?

I propose two living systems operating in sequence, at two different stages of the pipeline.

2.1 Stage 1. S. cerevisiae, as the hydrolysate feedstock. Food-grade yeast biomass is grown conventionally, then harvested and lysed by high-pressure homogenisation (HPH), followed by alkaline hydrolysis to break proteins into peptides and free amino acids (AAs), and tangential flow filtration (TFF) for size-based fractionation. What goes in: yeast cells, water, and mechanical/chemical energy. What comes out: a crude aqueous YH containing free AAs, peptides, and residual yeast cell-wall material, with an amino acid profile broadly similar in composition to that supplied by DMEM.

Image 2 - Overall diagram of lab-scale yeast-drived alternative basal media production process

2.2 Stage 2. Bovine primary satellite cells, engineered with SGN X, as the meat-producing chassis.

What SGN X aims to do: SGN X is a synthetic gene network I designed, as a conceptual exercise rather than a built construct, to solve a specific problem in myoblast expansion: satellite cells removed from the body and placed into culture are prone to drifting toward premature differentiation, which reduces the number of proliferative cells available before harvest. SGN X’s aim is to detect whether a cell is in the in vivo tissue environment or in ex vivo culture, and, whenever it senses ex vivo culture conditions, to actively suppress the cell’s ability to differentiate, while also including a fail-safe that eliminates any cell that differentiates anyway.

Image 3 - Synthetic Gene Circuit (SGN X)

The circuit’s components: SGN X consists of four functional cassettes, built around one sensing input and three response modules.

Fragment F (the sensor): an oxygen-gated switch. It contains a hypoxia response element (HRE) promoter driving a tetracycline repressor (TetR), alongside a separate TetO operator-minCMV promoter driving a bicistronic output of two synthetic transcription factors, TF1 and TF3, joined by a P2A self-cleaving peptide so that they are co-expressed then separated.

Fragment Y (used twice, as Y1 and Y2): a two-stage “detector and decision” module. Y1 is gated by TF1 and contains a MyoD-responsive promoter (built from E-box motifs) driving a second transcription factor, TF2. Y2 is gated by TF2 and drives two outputs: an epigenetic silencing cassette (a CpG methyltransferase paired with the methyl-CpG-binding factor CIBZ) to reinforce repression of myogenin (MyoG), and a separate AND-gated cassette that expresses the apoptotic effector Bax specifically when MyoG is high and paired box protein 7 (Pax7) is low.

Fragment M (the repressor): a CRISPR interference (CRISPRi) module driven by TF3, comprising catalytically dead Cas9 (dCas9) fused to KRAB-MeCP2 repressor domains, together with a U6-driven single-guide RNA (sgRNA) that targets regulatory DNA near the MyoD promoter.

How the mechanism works: The satellite cell niche in vivo is relatively low in oxygen (hypoxic), while cells expanding ex vivo in culture flasks are exposed to much higher, normoxic oxygen levels. Fragment F is built to read this difference. Under hypoxia, the stabilised hypoxia-inducible factor-1 alpha (HIF-1α) activates the HRE promoter, switching TetR on; TetR then binds TetO and represses the minCMV output, so TF1 and TF3 stay off, and the rest of the circuit is silent. Under normoxia, by contrast, HIF-1α is degraded, the HRE promoter is inactive, TetR is not made, and the TetO-minCMV cassette is released, so TF1 and TF3 are both expressed. From there, TF1 activates Y1, which senses MyoD activity (the cell’s own commitment signal toward differentiation) and, if that signal is elevated, produces TF2 to activate Y2, whose two outputs reinforce a stem-like, non-differentiated state and trigger apoptosis in any cell that reaches a late, committed differentiation state anyway. In parallel, TF3 activates Fragment M, whose CRISPRi complex is recruited by the sgRNA to the MyoD promoter and represses MyoD transcription directly. Together, the sensor-to-repressor path (F to M) and the sensor-to-detector-to-safety path (F to Y1 to Y2) mean that, for as long as a cell registers itself as being in ex vivo culture, the circuit both directly suppresses the master differentiation switch (MyoD) and monitors for, and removes, any cell that slips past that suppression.

What comes out: an expanded pool of myoblasts, held in a proliferative state by the circuit, which can then be released into differentiation (for example by shifting culture conditions) to fuse into multinucleated myotubes and build muscle-like tissue at the point of harvest.

Assumptions I’m making explicit: SGN X was designed as a conceptual exercise using notional genetic elements, not built or tested experimentally, and its oxygen-gated logic was extrapolated from evidence in other cell types rather than validated in bovine satellite cells directly. My hydrolysate growth data used C2C12 (an immortalised mouse myoblast line) as a proxy, not primary bovine cells. I am assuming both the medium effect and the circuit’s behaviour generalise to the actual target chassis, which is unverified in either case.

3. How would I use automation?

I would automate the medium formulation and screening step using a desktop liquid-handling robot such as an Opentrons.

In my own bench-scale MRes work, screening combinations of variables, including hydrolysate type (whole vs. size-fractionated), replacement percentage, BM background level, and osmolality-correction volumes, was done manually, at limited replication, across multi-well plates, constraining both the number of conditions tested and the statistical confidence of the results.

An Opentrons-class robot could automate the liquid-handling steps of this workflow: precise, repeatable pipetting of hydrolysate fractions, BM, phosphate-buffered saline (PBS, used for osmolality correction), and serum into defined well combinations across a full formulation matrix, at much higher throughput than manual pipetting. This is the natural automation target for three reasons: it is the most labour-intensive and error-prone manual step in the current process; it is highly parallelisable; and it is the step most directly limiting statistical rigor, since moving to higher replicate numbers becomes trivial once pipetting is automated. Downstream, this would pair with automated image-based cell counting for readout, and, specifically for SGN X, automated fluorescence-based screening (if a reporter were added to the circuit) to track what fraction of cells remain in the undifferentiated state across formulation conditions, closing the loop between medium composition, circuit performance, and proliferative yield.

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

4.1 How much?

I’ll scale to a modest but meaningful benchmark: enough cultivated meat to replace 1% of annual UK minced beef consumption.

UK beef consumption is roughly 1.1 million tonnes/year (rough estimate; UK per-capita beef consumption is on the order of 15 kg/year, times around 67 million people, giving approximately 1 million tonnes). 1% of this equals around 10,000 tonnes of cultivated muscle mass/year. From my own lab-scale process, a single hydrolysate production batch (a small, defined mass of yeast in a litre-scale volume of water) yields enough material to support on the order of a few litres of hydrolysate-supplemented medium, at the replacement level that supported cell growth in my own screening. Bioreactor cell culture for cultivated meat is typically estimated to require on the order of 500 to 2,000 L of medium per tonne of final product, a commonly cited industry planning figure. Taking a working estimate of approximately 1,000 L medium per tonne product: 10,000 tonnes times 1,000 L/tonne gives 10,000,000 L (10 million litres) of medium/year Scaling up from my lab-scale batch yield, this implies on the order of one to a few million hydrolysate production batches/year at lab-scale yeast loading, corresponding to a total yeast feedstock requirement in the low tens of tonnes/year

This is small relative to existing industrial yeast fermentation capacity, suggesting feedstock availability is not the binding constraint; cell-culture scale-up is.

4.2 How long?

Two different automation regimes apply here, as flagged in the assignment prompt.

Prototyping/screening scale (Opentrons): A single Opentrons run can prepare a 96-well formulation screen, covering dozens of hydrolysate-fraction, replacement, and background combinations, and, with a fluorescent SGN X reporter, circuit-performance readouts across those same conditions, in roughly 1 to 2 hours of robot time, versus a full day of manual pipetting for an equivalent plate. Over a proliferation assay window on the order of several days, matching the timeframe used in my own screening, a full design-of-experiments screen could be completed in a single week-long cycle rather than several weeks.

Manufacturing scale (industrial bioreactors and continuous TFF/HPH lines): HPH and TFF are both continuous, high-throughput industrial unit operations already used at scale in biopharma. Assuming an industrial HPH/TFF line processing approximately 1,000 L/hour of yeast slurry, producing the roughly tens of tonnes/year of hydrolysate feedstock needed above (equivalent to roughly 1 to 2 million litres of dilute, lab-concentration yeast slurry) would take:

On the order of 1,000 to 2,000 hours, or roughly two to three months of continuous operation (much less with parallel lines or higher yeast loading; for example, on the order of a few days at industrially typical yeast loading run continuously)

The cell expansion step is the actual long pole: bovine satellite cells at realistic doubling times (around 24 to 48 hours) require multiple weeks of bioreactor residence time per production cycle to reach meaningful biomass, independent of medium supply. This means the hydrolysate supply chain, once at industrial throughput, would not be rate-limiting; cell proliferation kinetics, bioreactor capacity, and, for the SGN X component, the efficiency of the differentiation-suppression circuit would be.

5. What don’t I know yet?

Whether the hydrolysate growth effect, and SGN X’s oxygen-gated logic, actually transfer from the immortalised mouse C2C12 proxy to primary bovine satellite cells. My hydrolysate data, which showed a clear dose-dependence between lower and higher replacement levels along with a benefit from diluting the basal medium with PBS, and SGN X’s design, extrapolated from HIF-1α/MyoD biology studied mostly in other species and contexts, both need direct testing in the actual target chassis before any of the numbers above, or the circuit’s assumed behaviour, can be trusted. Whether the hydrolysate production process still works, and still supports growth, once yeast loading is scaled from lab-scale, dilute conditions to industrially realistic, much higher loading. Higher yeast loading is the single biggest lever for improving amino-acid-per-litre efficiency, but likely also changes osmolality, protein recovery, and the concentration of inhibitory yeast cell-wall components; this is untested outside the dilute lab-scale regime. Whether SGN X’s fail-safe logic is tight enough to avoid false-positive apoptosis, or false-negative escape from the differentiation-suppressed state, at production scale. The circuit was designed conceptually, not built, and depends on several assumed thresholds (oxygen tension cutoffs, relative TetR/TetO repression strength, the MyoG-high/Pax7-low AND-gate logic for the Bax fail-safe) that would need to be experimentally tuned. An imperfect circuit could either fail to suppress differentiation, reducing proliferative yield, or trigger unwanted cell death across a bioreactor-scale population.

AI use: I used Claude (Anthropic) to help structure this proposal against the assignment’s four-question framework, to integrate material from my own unpublished MRes research and an earlier coursework essay (both my own prior work) into that structure, and to sanity-check the back-of-envelope scaling calculations in Section 4.

Final Project