Week 1 HW: Final Project Ideas

(1) What I Would Like to Biomanufacture and Why

Target product: A freeze-dried, paper-embedded cell-free diagnostic reagent that produces a visible colorimetric signal in the presence of Streptococcus agalactiae (Group B Streptococcus, GBS) nucleic acid from a vaginal or rectal swab collected at 35–37 weeks of gestation.

To be precise about what is being manufactured here: the product is not a drug or a biomaterial but a sensing reagent - a biological system preserved on paper, distributed to a clinic, and activated by patient sample material. The living system’s job is to make that reagent, not to operate inside a patient in real time. This distinction matters, and it is why a cell-free biomanufacturing approach is particularly well-matched to this application.

The Clinical Problem

GBS (Streptococcus agalactiae) is the leading infectious cause of neonatal sepsis, pneumonia, and meningitis in the first week of life. Global estimates place the annual burden at approximately 390,000 neonatal GBS infections, resulting in roughly 90,000 infant deaths, with the highest mortality concentrated in sub-Saharan Africa and South Asia.[1,2] Low- and middle-income countries (LMICs) account for approximately 99% of neonatal sepsis-related deaths globally.[3]

The mechanism of transmission is well understood. Approximately 18% of pregnant women worldwide carry GBS asymptomatically in the rectovaginal tract, with regional estimates ranging from 11% to 35%.[4] Vertical transmission occurs during labor and delivery. In well-resourced health systems, this is managed through a two-step protocol: a vaginal and rectal swab is taken at 35–37 weeks of gestation, cultured in selective enrichment broth for 24–48 hours, and colonized women are given intrapartum antibiotic prophylaxis (IAP) during labor. This strategy, when implemented with high coverage, substantially reduces early-onset neonatal GBS disease.[5]

The problem is not that we lack an effective intervention. It is that the intervention depends entirely on laboratory infrastructure - selective media, trained microbiologists, incubators, result-reporting systems - that is simply not available in the settings carrying the highest disease burden. Where routine culture-based screening cannot be implemented, colonized mothers are not identified, IAP is not given, and neonates die from a preventable infection.

A point-of-care GBS diagnostic - one that requires no laboratory, operates at room temperature, and gives a readable result in under an hour - would directly enable IAP in settings that currently have no pathway to screen. Existing rapid antigen tests lack sensitivity, and current nucleic acid amplification tests (NAATs), while sensitive, require stable electricity, trained operators, and expensive equipment. A freeze-dried, paper-based cell-free biosensor represents a distinct and currently unoccupied position in the diagnostic landscape.

My interest in this project is also personal. Maternal and neonatal health is the area of medicine I find most compelling, and the idea that a straightforward screening test - if made accessible - could prevent tens of thousands of deaths per year from a known, treatable infection is a powerful motivation.

(2) What Living System I Would Use

Chassis: Escherichia coli, used in two distinct capacities.

Overview of the Manufacturing and Sensing Logic

The living system here is not the diagnostic itself - it is the factory that produces the diagnostic reagent. The key product of biomanufacturing is cell-free transcription-translation (TX-TL) extract: a clarified lysate of E. coli that retains all the molecular machinery necessary for in vitro gene expression (ribosomes, RNA polymerases, elongation factors, aminoacyl-tRNA synthetases, energy regeneration systems) but contains no living cells and poses no biosafety concerns in deployment.

The diagnostic circuit embedded in that extract is a toehold switch: a synthetic mRNA with an engineered secondary structure that sequesters the ribosome binding site and start codon in a stable hairpin. The hairpin only unfolds when a complementary trigger RNA sequence - derived from the patient’s sample - binds to it. This allows ribosomes to access the start codon and translate a reporter protein (such as catechol 2,3-dioxygenase, which generates a visible yellow color, or beta-galactosidase with a chromogenic substrate). No signal is produced in the absence of the correct trigger sequence; translation, and therefore color change, only occurs when GBS-specific RNA is present.[6]

System components:

Toehold Switch Design

Candidate switch trigger sequences will be designed computationally using NUPACK for secondary structure prediction, combined with BLAST screening against commensal flora genomes, to maximize on-target activation by GBS 16S rRNA and minimize off-target activation by non-GBS Streptococcus species.[6] A library of candidate sequences targeting different hypervariable regions of the GBS 16S rRNA will be generated in silico, ranked by predicted specificity and fold-change upon trigger binding, and the top candidates taken into experimental screening.

What Goes In, What Comes Out

• Input: A vaginal/rectal swab at 35–37 weeks; a simple lysis buffer and co-lyophilized NASBA reagents on the paper strip; a 41°C heat source (heat block or warm water bath).
• Biological process: NASBA amplifies GBS 16S rRNA from swab material; amplified RNA trigger binds and unfolds the toehold switch; TX-TL extract translates the reporter gene; reporter enzyme converts a co-spotted chromogenic substrate.
• Output: A visible colorimetric signal readable by eye within 30–60 minutes; no laboratory instruments required.

Assumption: I am assuming that NASBA and TX-TL components can be co-lyophilized on the same paper strip, or on adjacent sequential zones. The compatibility of NASBA enzymes with TX-TL extract lyophilization has not been demonstrated for this specific combination and is a key open experimental question.

(3) How I Would Use Automation

Primary automation platform (R&D phase): Liquid-handling robot (Opentrons OT-2 or Hamilton STAR) operating in a Design-Build-Test-Learn (DBTL) loop.

The Prototyping Problem

The central bottleneck for this biosensor is not extract production - that step is fast and high-yielding. The bottleneck is switch sequence design and specificity validation. To determine which toehold switch sequences are both sensitive enough to detect clinically relevant GBS titers and specific enough not to activate on commensal flora, dozens to hundreds of candidate sequences must be tested in parallel. This is exactly the kind of high-dimensional screening problem that liquid-handling automation is designed to solve.

Automated DBTL Loop

[Step 1 - Design]
Computational pipeline (NUPACK + BLAST) generates a ranked list
of candidate toehold switch sequences targeting GBS 16S rRNA

             |

[Step 2 - Build]
Synthesized switch constructs prepared as in vitro transcription
templates; liquid-handling robot dispenses TX-TL extract + switch
construct + NASBA-amplified trigger RNA into 384-well plates

             |

[Step 3 - Test]
Each well challenged with:
  (a) Synthetic GBS trigger RNA at varying concentrations
      --> true-positive dose-response panel
  (b) Non-GBS Streptococcus trigger RNAs
      --> cross-reactivity panel
  (c) Buffer-only control
      --> background / false-positive rate

Fluorescence (GFP reporter) or colorimetric absorbance (420 nm)
kinetics read on plate reader over 60 minutes

             |

[Step 4 - Learn]
Signal-to-background ratios, dose-response curves, and cross-reactivity
scores fed back into computational model; next set of switch variants
prioritized for synthesis and re-screening

A single 384-well plate screens approximately 90 switch variants against 4 conditions in one day. Without automation, the same number of variants would require weeks of manual pipetting and introduce significant operator-to-operator variability. The DBTL loop can complete 3–5 full cycles within a month, converging on high-specificity candidates far more rapidly than sequential manual testing.

Manufacturing-Scale Automation

The R&D automation described above is for switch optimization. At manufacturing scale, the bottleneck shifts entirely:

1. Extract production: Large-scale E. coli fermentation in 100 L bioreactors with automated dissolved oxygen (DO), pH, and feed control.
2. Lyophilization throughput: Industrial freeze-dryers processing thousands of paper disks per batch; automated reagent-spotting systems analogous to lateral-flow strip manufacturing lines.
3. Batch quality control: Automated plate-reader QC testing of every production batch against GBS-positive and GBS-negative reference standards.

These are two fundamentally different automation contexts. The Opentrons optimizes a switch in a university lab; producing 6 million test strips per year requires purpose-built manufacturing infrastructure that does not yet exist for this specific product format. This is one of the key non-biological scale-up challenges the project would need to address.

(4) How Much Would I Need to Make, and How Long Would It Take?

(a) How Much?

Scale definition: Enough cell-free extract reagent to support annual antenatal GBS screening for 5 million pregnant women in settings that currently lack routine culture-based screening - primarily sub-Saharan Africa, South Asia, and parts of Southeast Asia.

There are approximately 135 million births globally per year.[8] The vast majority of pregnant women in LMICs receive no antenatal GBS screening; 5 million tests per year represents a meaningful initial deployment scale - comparable in ambition to how SARS-CoV-2 rapid antigen tests were initially rolled out in lower-income settings.

Scale calculation:

60 liters of cell-free extract per year is the manufacturing target for the bioproduction step.

Note: This calculation covers only the TX-TL extract component. NASBA enzymes (AMV-RT, RNase H, T7 RNA polymerase) and nucleotide substrates would be sourced separately from commercial suppliers or produced in a parallel process and are not included here.

(b) How Long?

Extract production rate:

Standard E. coli TX-TL extract production yields approximately 1–3 mL of active lysate per liter of culture using shake-flask growth and sonication. Optimized high-density fermentation protocols report 5–10 mL of active lysate per liter of culture.[10]

Conservative estimate (2 mL/L, standard protocol):

Extract needed: 60,000 mL
At 2 mL/L yield --> 30,000 L of E. coli culture required per year

In a 10 L benchtop bioreactor:
30,000 L / 10 L per run = 3,000 runs per year --> not feasible

This arithmetic reveals an important constraint. At low extract yield, the volume of culture required is prohibitively large for a small bioreactor.

Optimized estimate (10 mL/L, high-density fermentation):

Extract needed: 60,000 mL
At 10 mL/L yield --> 6,000 L of E. coli culture required per year

In a 100 L pilot-scale bioreactor (48 h per batch including growth, lysis, clarification):
Batches available per year: 365 days / 2 days per batch = ~182 batches
Batches needed: 6,000 L / 100 L = 60 batches per year (~1 batch per week)

--> Annual demand met with 1 batch per week in a single 100 L bioreactor.

At optimized extract yield, a single pilot-scale bioreactor running approximately one batch per week comfortably covers annual demand for 5 million tests. This is well within the capacity of a contract manufacturing organization running cell-free extract production as a dedicated process.

Project timeline estimate:

The biomanufacturing step itself is not the rate-limiting step for this project. The real bottlenecks are specificity validation, field stability testing, and regulatory pathway - none of which are primarily biological, and all of which take longer than the fermentation process they depend on.

(5) What I Don’t Know Yet

1. Will a toehold switch trigger sequence for GBS achieve sufficient specificity against commensal vaginal and rectal Streptococci?

This is the single biggest technical risk of the entire project. The vaginal and rectal microbiome contains multiple Streptococcus species that share sequence similarity with S. agalactiae, particularly in conserved regions of the 16S rRNA gene. Designing a switch that activates robustly on GBS 16S rRNA but not on S. pyogenes, S. parasanguinis, S. anginosus, or S. salivarius requires identifying a sufficiently divergent hypervariable target region. Toehold switches can be highly sensitive to single-base mismatches - which is advantageous for specificity - but off-target activation on closely related sequences has been documented.[6] This cannot be resolved by computational design alone; it requires a dedicated wet-lab cross-reactivity screen using clinical isolates of commensal flora, not synthetic sequences.

2. Does the system perform reliably from a raw swab without a nucleic acid extraction step?

Published demonstrations of cell-free paper-based biosensors have largely used purified or synthetic RNA as the trigger input. Real clinical swabs contain mucus, proteins, and other biomolecules that can inhibit both NASBA amplification enzymes and the TX-TL reaction. Extraction-free isothermal assays for GBS have been attempted using simple chemical lysis buffers with bacteriolytic enzymes, but whether those lysis conditions are compatible with NASBA chemistry and downstream TX-TL has not been established.[11] Sample preparation is historically where field performance of point-of-care nucleic acid tests most commonly fails, and this question must be validated using actual clinical swabs, not buffer-spiked laboratory controls.

3. How stable is the lyophilized system under real field conditions in the intended deployment settings?

Published studies report room-temperature stability of up to one year for lyophilized cell-free systems using cryoprotectant excipients such as trehalose.[12] However, “room temperature” in a controlled laboratory (20–25°C) is a very different environment from a rural primary health center in sub-Saharan Africa or South Asia, where ambient temperatures may routinely exceed 35–40°C and humidity is high. Whether lyophilized TX-TL extract retains functional activity after weeks on a shelf under those conditions cannot be inferred from accelerated stability data alone. Real-world field simulation studies, and eventually a pilot deployment, are required to answer this question.

References

1. Seale AC, Koech AC, Sheppard AE, et al. Maternal colonization with Streptococcus agalactiae and associated stillbirth and neonatal disease in coastal Kenya. Nature Microbiology. 2016;1(7):16067.
2. Lawn JE, Blencowe H, Oza S, et al. Every Newborn: progress, priorities, and potential beyond survival. The Lancet. 2014;384(9938):189–205.
3. World Health Organization. Newborns: improving survival and well-being. WHO Fact Sheet. September 2020.
4. Kwatra G, Cunnington MC, Merrall E, et al. Prevalence of maternal colonisation with group B Streptococcus: a systematic review and meta-analysis. The Lancet Infectious Diseases. 2016;16(9):1076–1084.
5. Prevention of Group B Streptococcal Early-Onset Disease in Newborns. ACOG Committee Opinion No. 797. Obstetrics and Gynecology. 2020;135(2):e51–e72.
6. Green AA, Silver PA, Collins JJ, Yin P. Toehold switches: de-novo-designed regulators of gene expression. Cell. 2014;159(4):925–939.
7. Compton J. Nucleic acid sequence-based amplification. Nature. 1991;350(6313):91–92.
8. United Nations Population Division. World Population Prospects 2022. New York: United Nations; 2022.
9. Pardee K, Green AA, Takahashi MK, et al. Rapid, Low-Cost Detection of Zika Virus Using Programmable Biomolecular Components. Cell. 2016;165(5):1255–1266.
10. Silverman AD, Kelley-Loughnane N, Lucks JB, Jewett MC. Deconstructing Cell-Free Extract Preparation for in Vitro Activation of Transcriptional Genetic Circuitry. ACS Synthetic Biology. 2019;8(2):403–414.
11. Gouverneur M, et al. Direct detection of Group B Streptococcus from vaginal swabs using loop-mediated isothermal amplification without prior nucleic acid extraction. Journal of Clinical Microbiology. 2021.
12. Pardee K, Slomovic S, Nguyen PQ, et al. Portable, On-Demand Biomolecular Manufacturing. Cell. 2016;167(1):248–259.e12.