Synthetic insulin is made by inserting the human gene for insulin into bacteria or yeast, which then produce the protein in large fermentation tanks. The finished product is purified, folded into its correct shape, and tested to meet strict pharmaceutical standards. This process, called recombinant DNA technology, replaced the old method of extracting insulin from animal pancreases and now supplies virtually all of the world’s insulin.
Why Animal Insulin Was Replaced
For decades after insulin’s discovery in the 1920s, every vial came from the pancreases of cows and pigs. The scale was staggering: by 1981, producing a single pound of insulin required 8,000 pounds of glands from roughly 23,500 animals, and that was only enough to treat 750 people for one year. Throughout the 1970s, the U.S. supply of pancreas glands shrank while meat prices swung unpredictably, making the entire system fragile and expensive.
On October 28, 1982, the FDA approved Humulin, the first medication ever made using recombinant DNA technology. It was identical to human insulin at the molecular level, and it freed production from its dependence on slaughterhouses. Today, human insulin has an estimated manufacturing cost of roughly $2.28 to $3.37 per 1,000-unit vial.
Step 1: Building the Gene
Human insulin is a small protein made of two short chains of amino acids, called the A chain and the B chain. To manufacture it, scientists first need a DNA sequence that codes for these chains. In the original Genentech process, the genes for the A and B chains were chemically synthesized in the lab and inserted separately into two different strains of E. coli bacteria. Each strain produced one chain, and the chains were later combined.
Modern production typically uses a single gene that codes for proinsulin, a longer precursor molecule that contains both chains connected by a short linking segment. This simplifies the process because the cell produces one protein instead of two, and the linking segment is trimmed away later during purification.
Step 2: Inserting DNA Into Host Cells
The synthetic insulin gene is spliced into a small, circular piece of DNA called a plasmid. Think of the plasmid as a delivery vehicle: it carries the insulin gene into the host cell and instructs the cell to read it. Once inside, the cell treats the insulin gene like one of its own and begins producing insulin protein every time it divides.
Two organisms dominate commercial insulin production. E. coli bacteria grow quickly, need simple nutrients, and produce high yields at low cost. Their main drawback is that the insulin protein often clumps into dense masses called inclusion bodies inside the cell, which then need to be dissolved and refolded into the correct shape. Yeast (Saccharomyces cerevisiae) can secrete the insulin precursor outside the cell, which simplifies collection. Yeast is also easy to scale up in large bioreactors. Novo Nordisk, for example, uses yeast for much of its insulin production, while Eli Lilly historically relied on E. coli.
Step 3: Fermentation
Once a strain of engineered bacteria or yeast is ready, it’s grown in large stainless-steel fermentation tanks that can hold thousands of liters. The cells are fed a nutrient broth and kept at carefully controlled temperature and pH. As they multiply, they churn out insulin protein (or proinsulin) continuously. A chemical or molecular signal triggers the cells to ramp up insulin production at the right moment in the growth cycle.
After fermentation, the mixture is a soup of cells, cell debris, nutrients, and insulin protein. Everything that isn’t insulin needs to be removed.
Step 4: Harvesting and Refolding
If the host organism is E. coli, cells are broken open to release the insulin protein trapped inside. The protein at this stage is misfolded, tangled up in inclusion bodies. Workers dissolve these clumps using chemical agents, then carefully guide the protein into its correct three-dimensional shape. This is critical because insulin’s two chains must be linked by three specific bonds between sulfur-containing amino acids (called disulfide bonds). If even one bond forms in the wrong place, the insulin won’t work.
Getting these bonds right is one of the trickiest parts of the entire process. Early methods yielded only about 25 to 30 percent correctly folded insulin, though optimized techniques using specialized chemical treatments have pushed yields to 30 to 60 percent. The approach relies on the fact that the correct fold is the most thermodynamically stable shape, so under the right chemical conditions, the protein naturally gravitates toward it.
If the host is yeast, the cells secrete a proinsulin precursor into the surrounding liquid, so there’s no need to crack cells open. The connecting segment between the A and B chains is then cut away using enzymes, leaving mature insulin.
Step 5: Purification
Raw insulin from either host contains bacterial proteins, cell fragments, endotoxins, and misfolded insulin molecules. Purification runs the mixture through multiple rounds of chromatography, a technique that separates molecules based on size, charge, or how they interact with specific surfaces. Each pass removes a different category of contaminant.
The U.S. Pharmacopeia sets strict limits on what can remain in the final product. Proinsulin contamination must be no more than 10 nanograms per milligram of insulin. Large clumped proteins can’t exceed 1 percent. Bacterial endotoxins (fever-causing molecules shed by bacteria) are capped at 10 endotoxin units per milligram. Total bacterial counts must stay below 300 colony-forming units per gram. The finished insulin must deliver at least 27.0 units of activity per milligram to earn a “purified” label.
How Insulin Analogs Are Engineered
Standard synthetic insulin is identical to the insulin your pancreas makes. But manufacturers also produce modified versions called analogs, designed to act faster or last longer than natural insulin. These are made using the same recombinant process, with small, deliberate changes to the amino acid sequence of the gene before it goes into the host cell.
Rapid-acting analogs work by preventing insulin molecules from clumping together after injection. Natural insulin forms clusters of six molecules (hexamers) under the skin, which must break apart before insulin can enter the bloodstream. Small amino acid swaps discourage this clumping:
- Lispro (Humalog): Two amino acids near the end of the B chain swap positions, proline and lysine at positions 28 and 29.
- Aspart (Novolog): Proline at position B28 is replaced with aspartic acid.
- Glulisine (Apidra): Two substitutions, one at B3 and one at B29, further reduce self-association.
Long-acting analogs take the opposite approach, slowing absorption so the insulin works over 18 to 24 hours. Insulin glargine (Lantus) adds two arginine amino acids to the end of the B chain and swaps one amino acid on the A chain. These changes shift the protein’s solubility so it forms tiny crystals under the skin that dissolve gradually. These analogs cost roughly six times more than standard human insulin, largely because of pricing decisions rather than production complexity.
Biosimilar Insulin
As patents on brand-name insulin analogs expire, other manufacturers can produce biosimilars: versions that are highly similar to the original but made in a different facility, sometimes using a different cell line or purification setup. Because insulin is a biological product (not a simple chemical), biosimilars can’t be exact copies the way generic pills are. They must demonstrate through clinical testing that they perform the same way in the body.
Some biosimilars go a step further and earn an “interchangeable” designation from the FDA, meaning a pharmacist can substitute them for the brand-name product without needing the prescribing doctor’s approval. This distinction requires additional data showing that switching between the biosimilar and the original causes no difference in safety or effectiveness. Several interchangeable biosimilar insulins have reached the U.S. market, expanding access and putting downward pressure on prices.