Alvaro Hernandez Zelada, student at National University of San Marcos, interested in immunomodulators, immunology, cancer, nanoparticlesand protein denovo design.
See the nav bar on the left to look through my work for How To Biomanufacture Almost Anything (Fall 2026)!
Product: Folate-receptor-targeted lipid nanoparticles (FA-LNPs) co-encapsulating Cas9 mRNA (modified with N1-methylpseudouridine) and two single-guide RNAs — sgRNA-E6 and sgRNA-E7 — directed against the HPV16 oncogenes responsible for cervical cancer. The final product is a sterile, injectable nanoparticle suspension of approximately 100 nm formulated in PBS with 10% sucrose as cryoprotectant.
Cervical cancer kills approximately 342,000 women per year worldwide, with over 95% of cases caused by HPV and HPV16 alone driving 55–60% of them (WHO, 2023). The virus encodes two oncoproteins — E6 and E7 — that are the central engines of malignant transformation. E6 hijacks the E6AP ubiquitin ligase to degrade p53, eliminating the cell’s primary tumor suppressor. E7 binds and inactivates pRb, releasing E2F1 and driving uncontrolled proliferation. Because E6 and E7 are expressed exclusively in HPV-infected cells and are entirely absent from normal human tissue, they are ideal CRISPR targets: disrupting them would selectively force tumor cells into apoptosis without harming healthy tissue.
Current standard-of-care treatments — surgery, cisplatin, radiotherapy — are effective in early stages but fail in roughly 30–40% of advanced cases and carry significant morbidity. A CRISPR-based therapy permanently disrupting E6/E7 removes the viral molecular program that keeps the cell cancerous, restoring p53 and pRb function and triggering cell death through the tumor’s own intact machinery. The targeting strategy uses folate receptor alpha (FRα), which is overexpressed in ~72% of cervical tumors and is accessible to circulating particles in tumor vasculature, while being sequestered on the apical (non-vascular) surface in normal epithelium — providing a genuine therapeutic window.
The manufacturing chassis is a two-stage biological system: E. coli for plasmid amplification, followed by a cell-free in vitro transcription (IVT) system for RNA production. The FA-PEG-DSPE lipid component is produced separately by chemical synthesis (NHS-ester conjugation of folic acid to DSPE-PEG-NH2).
E. coli strain DH5α harboring plasmid pLCAS9 (11,734 bp) is grown overnight in LB + ampicillin. The bacterium replicates the plasmid autonomously using its own DNA replication machinery. Critically, it ignores every eukaryotic and phage promoter on the plasmid — SP6, U6, CMV — because none of them are recognized by bacterial sigma factors. The bacterium reads only its AmpR promoter to survive selection pressure. No Cas9 protein is produced at this stage.
What goes in: LB medium + ampicillin + transformed E. coli inoculum.
What happens: Plasmid replication to hundreds of copies per cell over 16–18 h.
What comes out: 1–5 mg of purified plasmid DNA per liter of culture after maxiprep (alkaline lysis + anion-exchange column), verified at A260/A280 > 1.8.
Assumption: The Cas9 gene in pLCAS9 is human codon-optimized — standard for any CRISPR vector designed for mammalian expression.
The purified plasmid is linearized with a restriction enzyme (HpaI, cutting downstream of the poly-A120 sequence) and used as a template for in vitro transcription — a fully cell-free enzymatic process. Three RNA molecules are produced in separate reactions:
What goes in: Linearized DNA template + RNA polymerase + NTPs (with m1Ψ for Cas9 mRNA) + ARCA cap analog + RNase inhibitor + buffer.
What happens: Polymerase binds the SP6/T7 promoter on the linearized template and synthesizes RNA 5’→3’. The physical end of the linearized DNA forces the polymerase to fall off, producing a transcript of defined length. Circular plasmid cannot be used because the polymerase has no stop signal and produces multi-kilobase aberrant readthrough transcripts.
What comes out: Purified, capped, polyadenylated, m1Ψ-modified Cas9 mRNA and two functional sgRNAs. Typical yield: 1–4 mg RNA per mL of IVT reaction. Verified by denaturing agarose gel, Nanodrop, and Bioanalyzer RNA chip.
The three RNAs are then co-encapsulated into the lipid nanoparticle at a Cas9 mRNA : sgRNA-E6 : sgRNA-E7 mass ratio of 3:1:1 using microfluidic mixing.
At lab and optimization scale, an Opentrons OT-2 liquid-handling robot automates the highest-variability steps:
The Opentrons is inadequate at scale. Two platforms replace it:
NanoAssemblr Ignite (Precision NanoSystems) uses a microfluidic herringbone mixing chip at controlled flow rates (12 mL/min, 3:1 aqueous:organic). It reduces batch-to-batch size variation from ±30–40 nm (manual) to ±5–10 nm, brings PDI consistently below 0.15, and scales from 0.5 mL to 250 mL per run.
Cloud Lab (Emerald Cloud Lab / Strateos) enables remote programming of hundreds of parallel IVT + formulation + characterization experiments for systematic Design of Experiment (DoE) optimization — for example, a fractional factorial design across lipid ratio, N/P ratio, PEGylation density, RNA cargo ratio, and mixing speed simultaneously. This is impossible to execute manually and would take months on a single benchtop.
Target: 10,000 patients per year with HPV16+ cervical cancer in Latin America (Peru, Bolivia, Paraguay — incidence ~20–30 per 100,000 women), representing a realistic Phase II/III clinical trial scale or first-year commercial launch.
Dose per patient:
Total RNA needed per year:
240 mg/patient × 10,000 patients × 1.2 overage factor (QC, batch failures, regulatory retention) = 2,880 g RNA/year
IVT volume required:
At conservative yield of 2 mg RNA/mL → 1,440 L IVT reaction/year
At optimistic industrial yield of 10 mg/mL → 288 L IVT reaction/year
LNP formulation volume:
At 0.2 mg RNA/mL final concentration → 14,400 L LNP suspension/year
Per patient course: 1,200 mL total (300 mL per injection) — within IV infusion range.
Prototype (Opentrons OT-2 + benchtop IVT):
~50–100 mg RNA/day → would take approximately 80 years to hit the annual target. Clearly not viable for scale, but appropriate for optimization and proof-of-concept.
Pilot scale (NanoAssemblr + 10 L IVT bioreactor):
~5–10 g RNA/day → 288–576 days to reach the annual target.
Manufacturing scale (50–100 L IVT bioreactor + automated inline LNP line):
~50–100 g RNA/day → 29–58 days to produce enough for 10,000 patients.
Industrial scale (continuous 500 L IVT + inline microfluidic LNP):
~500 g RNA/day → ~6 days for the full annual batch.
The key insight is that the automation technology appropriate for prototyping (OT-2) is completely different from what manufacturing requires (continuous bioreactor IVT + inline microfluidics). The transition between scales is not incremental — it is a platform change.
Unknown #1 — In vivo editing efficiency in real tumors.
The 40–65% editing efficiency seen in SiHa cell culture may not translate to solid tumors in vivo. The tumor microenvironment is hypoxic, acidic, and immunosuppressive — all conditions that can degrade LNPs before they reach cancer cells. Additionally, FRα expression is heterogeneous within a tumor, meaning low-FRα cells could escape targeting entirely. What I would need to measure: LNP biodistribution and stability in tumor-bearing mice by IVIS imaging, FRα expression by single-cell RNA-seq in SiHa xenografts, and endosomal escape efficiency by galectin-8 recruitment assay.
Unknown #2 — Endosomal escape kinetics and Cas9 translation in the tumor context.
Only ~1–2% of endocytosed LNPs successfully escape the endosome in vivo. Whether this fraction is sufficient to produce enough Cas9 protein for therapeutic editing in an HPV16+ cervical tumor — where the microenvironment may upregulate innate immune sensors like TLR7/8 and MDA5 — is unknown. I would need a time-course Western blot of Cas9 protein in SiHa cells (2h through 72h), confocal immunofluorescence for nuclear localization, and a dose-response curve correlating intracellular Cas9 concentration with editing efficiency to establish the minimum effective dose.
Unknown #3 — IVT yield at manufacturing scale.
The entire scale calculation assumes 2–10 mg RNA/mL of IVT reaction. This holds for small reactions (20–100 µL) but degrades at large scale for three reasons: pyrophosphate accumulation (a polymerase inhibitor produced stoichiometrically with every NTP incorporated, reaching inhibitory concentrations above ~30 mM), heat inactivation of the polymerase in large vessels with inadequate mixing, and dsRNA byproduct accumulation that requires HPLC purification to remove — which becomes the rate-limiting step at liter scale. If the true yield is 5–10× lower than assumed, the manufacturing volumes and timelines calculated above are proportionally larger. I would need to characterize IVT yield as a function of reaction volume (1 mL → 1 L) with and without pyrophosphatase supplementation, and establish HPLC dsRNA removal capacity at liter scale.