Scientists today make insulin by inserting the human insulin gene into living cells, usually bacteria or yeast, and letting those organisms produce the protein in large fermentation tanks. This process, called recombinant DNA technology, replaced the older method of extracting insulin from pig and cow pancreases. The first product made this way, Humulin, was approved by the FDA on October 28, 1982, making it the first recombinant DNA medical product ever sold.
Building the Gene
Human insulin is a small protein made of two short chains: an A chain (21 amino acids) and a B chain (30 amino acids). To produce it artificially, scientists first need a DNA sequence that codes for these chains. In early work at Genentech and City of Hope in 1978, researchers chemically synthesized the genes for each chain from scratch, assembling short fragments of DNA and stitching them together. The synthetic gene for the B chain was 104 base pairs long; the A chain gene was 77 base pairs.
Today, gene synthesis is routine and automated. Scientists can order custom DNA sequences from commercial suppliers, specifying exactly which version of the insulin gene they want. This is especially useful for making insulin analogs, where the gene is deliberately altered to change one or two amino acids in the final protein.
Inserting the Gene Into a Host Cell
Once the insulin gene is ready, scientists splice it into a small, circular piece of DNA called a plasmid. Think of a plasmid as a delivery vehicle: it can slip inside a bacterial or yeast cell and instruct that cell to read the insulin gene and build the protein. The modified plasmid is mixed with the host cells under conditions that encourage them to take it up, and then those cells are grown in culture to confirm the gene is working.
Bacteria vs. Yeast as Cell Factories
The two most common host organisms are the bacterium E. coli and baker’s yeast (Saccharomyces cerevisiae). Each has trade-offs.
E. coli grows fast and is well understood, but it can struggle with certain genes. If the genetic “spelling” of the human insulin gene doesn’t match the patterns E. coli prefers, the bacterium produces very little protein or makes errors. Scientists work around this by rewriting the gene using synonymous codons that E. coli reads more efficiently, without changing the final protein. Another limitation is that E. coli often produces insulin in clumps of misfolded protein called inclusion bodies, which then have to be dissolved and refolded in the lab before the insulin is usable.
Yeast is a more complex cell, closer to human cells in its internal machinery. It can fold proteins more naturally and even add certain chemical modifications that bacteria cannot. Yeast-based systems have been reported to produce around 80 milligrams of insulin precursor per liter of culture. Novo Nordisk, one of the world’s largest insulin manufacturers, has long used yeast for its production lines.
Growing Cells at Industrial Scale
In a manufacturing facility, the modified cells are grown in large stainless-steel fermentation tanks called bioreactors. A typical research-scale batch might use 20 to 200 liters of culture broth, while commercial facilities operate at thousands of liters. The cells are fed nutrients in carefully controlled stages (a process called fed-batch fermentation) to maximize how much insulin they produce. From a 200-liter fermentation run using E. coli, manufacturers can recover roughly 138 grams of human insulin after processing.
Turning Proinsulin Into Active Insulin
The cells don’t make finished insulin directly. They produce a longer precursor called proinsulin, which contains the A chain and B chain connected by an extra segment called the C-peptide. In the human body, specialized enzymes in the pancreas snip out the C-peptide to activate the insulin. In the factory, scientists mimic this step using enzymes (typically trypsin and carboxypeptidase B, or engineered alternatives) that cut the proinsulin at precise locations, releasing the active two-chain insulin molecule and discarding the C-peptide.
Purification
After enzymatic conversion, the insulin is still mixed with bacterial proteins, DNA fragments, cell membrane debris, and leftover enzymes. Removing all of this requires multiple rounds of chromatography, a technique that separates molecules based on their physical and chemical properties.
The most critical step is reversed-phase chromatography, which sorts molecules by how water-repellent they are. Insulin binds to a column packed with hydrocarbon-coated beads (commonly C8 or C18 chains), and an organic solvent gradually washes it off, leaving impurities behind. This step is run at an acidic pH between 3.0 and 4.0, where insulin dissolves well. A second technique called size-exclusion chromatography separates molecules by size, filtering out any insulin that has clumped into larger aggregates. The final product must fall within 95% to 105% of its labeled potency, and high-molecular-weight impurities (clumped proteins) must stay below 1.7%.
Engineering Insulin Analogs
Standard recombinant human insulin is identical to what the pancreas makes, but it doesn’t perfectly mimic the body’s natural release patterns. To solve this, scientists engineer insulin analogs by swapping specific amino acids in the A or B chain. These small changes alter how quickly the insulin enters the bloodstream or how long it stays active.
Rapid-acting insulins are designed to work within minutes of injection. Insulin lispro, for example, has two amino acids near the end of the B chain swapped in position. Insulin aspart has a single change at position 28 of the B chain. These tweaks prevent insulin molecules from sticking together, so they absorb faster from the injection site.
Long-acting insulins take the opposite approach. Insulin glargine has two extra amino acids added to the end of the B chain and one substitution on the A chain, shifting its chemical properties so it forms tiny crystals under the skin that dissolve slowly over 24 hours. Insulin detemir has a fatty acid chain attached to it, which causes it to bind to a protein in the blood (albumin) and release gradually. The newest ultra-long-acting insulin, icodec, is engineered with multiple substitutions that weaken its binding to the insulin receptor and strengthen its binding to albumin, creating a once-weekly insulin.
All of these analogs are made using the same recombinant DNA process. The only difference is the starting gene sequence.
Stem Cell Approaches on the Horizon
A fundamentally different strategy is moving through clinical trials: growing insulin-producing cells from stem cells and transplanting them into patients. Rather than manufacturing insulin in a factory, this approach aims to restore the body’s own ability to make it. In a recent trial published in the New England Journal of Medicine, 14 people with type 1 diabetes received infusions of lab-grown islet cells derived from stem cells. All 14 showed signs of insulin production afterward, and 10 of 12 participants in the main study groups were completely off insulin injections at one year. Those participants spent more than 70% of their time in a healthy blood sugar range.
The treatment still requires immune-suppressing drugs to prevent rejection, and serious side effects occurred in the trial, including two deaths from complications related to immunosuppression. This technology remains experimental, but it represents the first time scientists have been able to manufacture functional insulin-producing cells outside the body and successfully transplant them.