01Executive summary
Three milestones that turned living systems into a manufacturing platform, then made the instructions programmable, then compressed the development timeline to months.
Recombinant human insulin, produced by engineered bacteria, was approved in 1982 — the first therapeutic protein made by a living cell line designed for the purpose. CRISPR-Cas9 was shown in 2012 to work as a guided cutting tool, making targeted edits routine. In 2020 an mRNA product with lipid nanoparticle delivery was manufactured and distributed in enormous quantity within a year of the target sequence becoming available.
02Recombinant production: the process defines the product
Before 1982, insulin was extracted from animal pancreases — supply limited by slaughterhouse throughput, with the product not quite identical to the human protein. Recombinant production inserts the human gene into a microorganism, which then manufactures the protein in culture. Supply becomes a fermentation capacity question rather than an agricultural one.
The engineering distinction from chemical manufacture is fundamental. A small-molecule drug can be characterised completely: its structure is known and any competent synthesis route producing that structure produces the same substance. A therapeutic protein is large, folded, and often modified after translation. Its properties depend on how the cells were grown, and analysis cannot fully characterise it. This is the origin of the principle that in biologics manufacture, the process defines the product.
- Cell line developmentA production line is engineered and selected for yield and stability, then banked. The bank is the master reference for everything made afterwards.
- Upstream cultureCells are grown in bioreactors under closely held conditions of temperature, pH, dissolved oxygen and feeding. Deviations change the product, not just the yield.
- Harvest and clarificationCells and debris are separated from the product stream without damaging a fragile molecule.
- PurificationSuccessive chromatography steps remove host proteins, nucleic acids and product variants. This dominates cost of goods.
- Formulation and fillThe purified protein is stabilised, filled aseptically and packaged. Sterility assurance is absolute rather than statistical in intent.
Scaling a bioreactor changes the environment cells experience even when every setpoint is unchanged. Mixing time lengthens, so cells traverse gradients in oxygen and nutrient concentration. Shear from impellers rises. Gas transfer, heat removal and carbon dioxide stripping all scale differently from volume. Cells respond to the environment they actually encounter, so a process that works at laboratory scale can produce a subtly different product at manufacturing scale. Scale is a process change requiring comparability evidence, not merely a bigger vessel — and this is the same lesson as heat of hydration at Hoover Dam, in a completely different discipline.
03Gene editing: targeting became programmable
Editing genomes was possible before 2012, using engineered proteins designed to recognise a specific sequence. The difficulty was that each new target required designing and validating a new protein — months of specialist work per target.
CRISPR-Cas9 separates the cutting function from the targeting function. A nuclease does the cutting and is unchanged between applications. Targeting is supplied by a short guide RNA that finds its site by base pairing. Changing target means synthesising a different short RNA, which is fast and cheap. The reduction in cost and time is the entire significance: a capability that existed became accessible.
Separate the general from the specific
A fixed general mechanism plus a cheap configurable element beats redesigning the mechanism each time. This is the same argument as the microprocessor replacing custom logic, arriving in molecular biology.
Accessibility changes the field
When a technique moves from specialist to routine, the number of people applying it rises by orders of magnitude, and progress follows the participation rather than the capability.
Credit here is contested and was litigated extensively. Francisco Mojica characterised CRISPR sequences and proposed their biological role. Rodolphe Barrangou and colleagues demonstrated the adaptive immune function experimentally. Emmanuelle Charpentier and Jennifer Doudna published the programmable guided cutting mechanism in 2012 and shared the Nobel Prize in Chemistry in 2020. Feng Zhang’s group demonstrated editing in eukaryotic cells, and patent proceedings over that application ran for years. This series names the contributions rather than nominating an inventor, which is the only defensible treatment.
The engineering limitations that matter
Off-target cutting at sequences similar to the intended one is the central technical risk, and detecting it reliably is harder than making the edit. Delivery into the right cells in a living organism, rather than into cells in a dish, remains the practical constraint for most therapeutic applications. And the repair after cutting is performed by the cell’s own machinery, so the outcome is not fully determined by the tool. Descriptions of gene editing as precise should be read with all three qualifications attached.
04mRNA at scale: a platform, and a cold chain
A conventional vaccine requires growing or synthesising the antigen itself, and the manufacturing process is largely specific to that antigen. An mRNA product instead delivers instructions, and the recipient’s own cells produce the antigen. The manufacturing process is substantially the same regardless of what the sequence codes for.
That is a platform in the strict sense used throughout this series: a fixed substrate with the product expressed as configuration. Changing target means changing a sequence, not redesigning a plant. It is why development compressed from years to months once the sequence was available.
| Problem | Resolution |
|---|---|
| mRNA provokes an innate immune response | Modified nucleosides reduce unwanted immune activation while preserving translation — work by Kariko and Weissman over many years |
| mRNA is rapidly degraded | Encapsulation in lipid nanoparticles protects the payload until it reaches cells |
| Naked mRNA does not enter cells efficiently | The same lipid nanoparticle mediates uptake, making the delivery vehicle as critical as the payload |
| Nanoparticle manufacture is not a batch mixing step | Controlled microfluidic or impingement mixing produces consistent particle size distribution; this proved a principal scale-up constraint |
| Stability in distribution | Frozen or ultra-cold storage, driving a distribution system built specifically for the product |
Manufacturing the product was one problem. Getting it to hundreds of millions of people at low temperature was a separate one, and it required freezers, insulated shipping with temperature logging, distribution routing designed around thaw windows, and dosing schedules that fit the stability data. A product that cannot be delivered is not a product. Whenever a technical achievement is being assessed, asking what the distribution and support system must look like usually reveals the harder half.
05Takeaways for current practice
- Where the process defines the product, control the process. Analysis cannot substitute for process control when the product cannot be fully characterised.
- Scale-up is a process change. Gradients, shear and transfer rates scale differently from volume, and comparability must be demonstrated.
- Separate the general mechanism from the specific configuration. It is the single most repeated high-value pattern in all three of these series.
- State the qualifications on a capability. Off-target effects, delivery and uncontrolled repair all bound what “precise editing” means.
- Design the distribution system as part of the product. The cold chain was as much an engineering achievement as the formulation.
Relevant references include ISO 13485 and the PIC/S Guide to Good Manufacturing Practice as adopted by the Therapeutic Goods Administration, ICH quality guidelines including Q8 to Q11, and AS/NZS 2243.3 for laboratory biological safety. Cited by number for orientation only — verify currency.