The sequence for the first covid vaccine took a weekend. The doses took most of a year, and the delay was not paperwork

The genetic sequence for the first covid mRNA vaccine was settled within days of the virus's genome being published. Actual doses in actual arms took most of a year. Almost every explanation offered for that gap points at trials and regulators, and trials genuinely took months — but underneath sat a manufacturing problem that had never been solved at that scale, and it is more interesting than the paperwork.

Insulin needs microbes. Antibodies need mammalian cells in bioreactors. An mRNA medicine needs neither: there are no living cells anywhere in the process. It is made by an enzyme, in a tank, from ingredients. That single fact makes the design almost trivially changeable and the production stubbornly difficult, and the two are connected.

A medicine assembled by an enzyme

The reaction is called in vitro transcription, and it is startlingly simple on paper. Put into a vessel: a DNA template carrying the sequence you want, an RNA polymerase borrowed from a bacteriophage, the four RNA building blocks, magnesium and buffer. The polymerase latches onto the template and spools out RNA copies. A few hours later you have grams of messenger RNA and the DNA template can be digested away.

Because there is no organism to keep alive, nothing about this process cares what the sequence says. Change the template and every other step is unchanged — same enzyme, same reagents, same equipment, same purification. This is the real content of the claim that mRNA is a "platform": the biology has been reduced to information, and information is cheap to change.

That is also why the vaccine could be designed in a weekend. Designing it was writing a sequence.

The parts that are not the sequence

What most people picture as the product — the instructions for one protein — is the least demanding part of the molecule. Around it sits architecture that determines whether anything works at all.

A cap is added at the front. Without it the cell's machinery treats the RNA as debris, refuses to translate it, and destroys it. A poly-A tail at the other end governs stability. Between them and the coding region sit untranslated regions whose sequence tunes how long the molecule survives inside a cell and how much protein it produces — chosen empirically, and a genuine source of competitive advantage between manufacturers.

Then there is the substitution that made the entire field possible. Our cells carry sensors that detect foreign RNA, because that is what a viral infection looks like from the inside. Ordinary synthetic mRNA trips them: the cell mounts an interferon response, shuts down translation, and you get inflammation instead of protein. Replacing one of the four bases with a chemically modified version dampens that recognition enough for the RNA to be read rather than attacked. That discovery is why therapeutic mRNA exists and why it earned a Nobel Prize in 2023 — a modification to a single building block, standing between an idea that did not work and one that did.

The impurity that decides everything

Here is the part that is genuinely under-explained, and it is where manufacturing difficulty concentrates.

The polymerase is not perfect. It produces truncated transcripts that stop early, and — more seriously — it produces double-stranded RNA by-products, when a transcript folds back on itself or the enzyme copies the wrong strand. These are a small fraction of the output and they are chemically almost the same substance as the product.

They are also exactly what the immune system is built to detect. The sensors that make single-stranded foreign RNA a problem are far more sensitive to double-stranded RNA, because double-stranded RNA is a near-unambiguous signature of viral replication. A batch carrying too much of it produces less protein than it should and more reactogenicity than it should, from the same nominal dose.

So a large part of mRNA manufacturing is a separation problem: removing a contaminant that differs from the product mainly in shape. This is done with chromatography steps chosen to exploit that difference, and how well it is done varies between manufacturers and between batches. Process changes that look minor can move the impurity profile, which is why an apparently trivial platform hides a considerable amount of accumulated know-how.

The sequence is information and can be changed overnight. The impurity profile is chemistry, and it cannot.

Wrapping it, where mixing becomes a manufacturing step

Naked mRNA would be destroyed in the bloodstream within minutes and could not cross into a cell in any case, so it has to be packaged — the delivery role lipid nanoparticles are known for. The part worth understanding is how that packaging is actually made, because it is not a chemical reaction.

Four lipids are used: an ionisable lipid that carries positive charge only in acid, a helper phospholipid, cholesterol for stability, and a polymer-tipped lipid that keeps particles from clumping. The lipids are dissolved in ethanol; the mRNA sits in acidic water. The two streams are then forced together very rapidly.

At that moment the ionisable lipid is charged, so it grabs the negatively charged mRNA, and as the ethanol dilutes, the whole assembly falls out of solution as particles — in milliseconds. Nobody builds the particle. It self-assembles, and the size and uniformity of what forms depend on how fast and how evenly the two streams met.

That makes mixing geometry a formulation parameter. Laboratories use microfluidic mixers; factories use jet impingement, where two streams collide in a small chamber. Scaling up an mRNA product is therefore substantially a fluid-dynamics exercise — reproducing a mixing environment at forty times the flow rate — rather than a chemistry one. After that the ethanol is removed and the buffer exchanged to neutral pH, which discharges the ionisable lipid and leaves the mRNA trapped inside.

Why the freezer requirements differed

The cold chain that shaped covid vaccine logistics was not a property of mRNA as such. Two things degrade these products: the RNA slowly hydrolyses, and the lipid particles oxidise and aggregate. Both are slowed by cold, and how much cold you need depends on the lipid composition, the buffer and the excipients — which is why two mRNA vaccines built on the same principle had visibly different storage requirements, and why later formulations improved on both. For a country distributing vaccines across a range of climates and a patchy cold chain, this is not a footnote; formulation is the difference between a dose that arrives usable and one that does not.

What the platform actually enables

The honest version of the platform argument is narrower than the marketing and more interesting.

Because only the template changes, the same facility can make a different product without being rebuilt — which matters most where the batch size is genuinely one. Personalised cancer vaccines sequence a patient's tumour, identify mutated proteins unique to it, and encode those in an mRNA made for that person. No other manufacturing modality treats a batch of one as normal.

The same delivery works for gene editing: an LNP carrying a base editor as transient instructions rather than as a permanent gene, which is the route around the one-shot problem that constrains viral gene therapy. And protein replacement is possible in principle, with the caveat that expression is deliberately temporary — excellent for a vaccine, awkward for a condition requiring a protein every day for life.

Why it matters for students and researchers

India manufactured covid vaccines at enormous scale using viral vector and inactivated virus platforms, and an Indian company developed and received emergency approval for a domestically made mRNA vaccine. But mRNA capability is not an extension of existing vaccine plants. It is a different discipline: cell-free enzymology rather than cell culture, RNA analytics rather than protein analytics, a lipid supply chain that barely existed before 2020, and scale-up dominated by mixing rather than by fermentation.

That is the useful observation for anyone entering the field. The three manufacturing modalities this site has now covered — microbial expression, mammalian cell culture, cell-free synthesis — look similar from a distance and share almost no skills. Each is defined by its own characteristic difficulty: refolding for microbial protein, glycosylation and cell-line stability for antibodies, impurity separation and mixing kinetics for mRNA. Knowing which difficulty a product carries tells you what a plant needs and what its people must be good at, and that is a more durable thing to understand than any individual process.

Frequently asked questions

How is mRNA for a vaccine actually made?

By in vitro transcription: a DNA template, an RNA polymerase, the four RNA building blocks and buffer are combined in a vessel, and the enzyme produces RNA copies. No living cells are involved at any stage.

Why is mRNA called a platform technology?

Because changing the product means changing only the DNA template. The enzyme, reagents, equipment and purification steps stay the same, so a facility can switch to a different target without being rebuilt.

What is the double-stranded RNA problem?

The polymerase produces small amounts of double-stranded RNA as a by-product, and the immune system detects that far more strongly than single-stranded RNA because it signals viral replication. It must be removed by chromatography, which is one of the hardest steps in the process.

Why do lipid nanoparticles need rapid mixing?

Because the particles self-assemble within milliseconds when a lipid-in-ethanol stream meets mRNA in acidic water. Particle size and uniformity depend on how fast and evenly the streams combine, so mixing geometry is a formulation parameter and a scale-up challenge.

Why did some mRNA vaccines need colder storage than others?

Because stability depends on the lipid composition, buffer and excipients rather than on mRNA alone. Different formulations degrade at different rates, which is why storage requirements varied between products and improved in later versions.