Peptide Origami: the Challenge of Insulin Folding

My drawing of the negative spaces in one of Dorothy Crowfoot Hodgkin’s models of insulin

We take it for granted that the 1-millimetre I measure with my ruler is the same as the 1-millimetre you measure with yours, yet creating new measurement standards are rarely as simple as this, least of all when it comes to biological measurements. I recently had the pleasure of visiting the National Physical Laboratory as part of the ASEAN-UK Regional Training and Workshop on Engineering Biology. NPL develops and maintains national primary measurement standards, and our visit coincided with the centenary of the standardisation of biological measurements – it is 100 years since the first international standard for insulin was adopted. Since then, insulin as a product has come a very long way.

In those days, insulin was purified from the pancreases of pigs or cows. Defining an international standard for insulin ensured that every single batch was of a quality good enough to be given to a patient. This helped to define what that quality was and provided a benchmark to which all batches of insulin would be compared.

Insulin serves a vital role in our bodies coordinating the uptake of sugar into cells to be used as energy or stored for later when it is needed. Blood sugar levels are highest after eating, and it’s important that this level is brought down before the sugar starts to cause damage to blood vessels and organs. In diabetes, insulin is either not made or is poorly recognised by the body, so diabetic patients must monitor their food intake and blood sugar carefully, injecting the exact amount of insulin that their body needs.

Almost sixty years after animal insulin was first used as a treatment for diabetes, scientists at Genentech developed a technique for using bacteria to make human insulin on a large scale [1]. Called Humulin, this started to be marketed by Eli Lilly in 1982, and was the first therapeutic to be made using DNA recombinant technology. This is a standard laboratory technique these days, yet the case of insulin shows how difficult the procedure can be if the protein you are trying to make doesn’t have an ideal chemistry.

Human insulin contains 51 amino acids split across two chains, A and B. These chains are linked by disulphide bridges, two connecting A and B, and a third connecting two parts of chain A. In my last post, on cone snail toxins, I wrote about how disulphide bonds give rigidity to peptides, likening a disulphide bond to a glued flap in a paper model, which can turn a piece of paper from 2D into 3D. However, as I mentioned in my post about using carnivorous plants to make recombinant proteins, one of the limitations of using E. coli lies in the difficulty it has in making complex proteins – specifically, those including disulphide bonds.

A disulphide bond is formed when two cysteine amino acids join together. Cysteine is first oxidised to cystine – this means that each cysteine amino acid gives an electron to oxygen, the terminal electron acceptor. “Here, hold this for me” – and oxygen is very obliging – only oxygen is not very nice if you give it an extra electron, turning it into a Reactive Oxygen Species. In this state, unstable oxygen causes lots of damage to the cell’s other components – and is why ‘antioxidants’ are marketed with such health benefits – antioxidants donate electrons to pair up single electrons into stable pairs.

As you can imagine, it is in a cell’s best interest to avoid letting oxygen get hold of unpaired electrons, and so all cells – including E. coli – pack their cytoplasms full of antioxidants like glutathione. This is great for preventing damage but becomes more problematic if you are wanting to use these bacteria to make a disulphide-bonded peptide like insulin.

Indeed, trying to use E.coli to make insulin results in large knots of peptides called inclusion bodies. The newly-made proteins need to be purified from these and then reconfigured to make the correctly wired peptides. In the original Genentech method [1], insulin chains are produced in separate E. coli strains, purified and mixed. The chemical conditions are then altered to untangle the peptides, then changed again to those which encourage the proper folding. This process, of purifying insulin from inclusion bodies, remains the standard practice for making insulin in bacteria today [2].

Some E. coli strains such as the Origami strains have been edited to inactivate the genes for glutathione reductase (gor) and thioredoxin reductase (trxB) enzymes, which otherwise ensure glutathione (and thioredoxin) are given enough electrons to freely donate for their antioxidant duties. Shuffle strains have been engineered to enhance disulphide bond formation with the addition of a cytoplasmic disulfide isomerase (DsbC) protein. It is possible to produce lispro insulin (an example of which is Humulin) using the Shuffle strain [3].

The difficulties in producing correctly-folded insulin naturally promotes the development of different methods, each of which can produce an insulin with slightly different properties, even if the insulin structure that is sought for is the same. Any variation in cell lines, media components, expression systems, bioreactor conditions – as well as the details of the purification steps themselves – can produce products that are of variable quality. Similarly, not all insulin is made in E. coli – Novo Nordisk use Saccharomyces cerevisiae – brewers’ yeast – to make their insulin (aspart insulin, brand name NovoRapid) leading to a product again with different characteristics.

References

1. Goeddel, D. V., Kleid, D. G., Bolivar, F., Heyneker, H. L., Yansura, D. G., Crea, R., Hirose, T., Kraszewski, A., Itakura, K. and Riggs, A. D. (1979) Expression in Escherichia coli of chemically synthesized genes for human insulin Proc Natl Acad Sci U S A. 76(1):106-110.

2. Zieliński, M., Romanik-Chruścielewska, A., Mikiewicz, D., Łukasiewicz, N., Sokołowska, I., Antosik, J., Sobolewska-Ruta, A., Bierczyńska-Krzysik,A., Zaleski, P. and Płucienniczak, A. (2019) Expression and purification of recombinant human insulin from E. coli 20 strain Protein Expr Purif 157:63-69.

3. Khalilvand, A.B., Aminzadeh, S., Sanati, M.H. and Mahboudi, F. (2022) Media optimization for SHuffle T7 Escherichia coli expressing SUMO-Lispro proinsulin by response surface methodology BMC Biotechnol 22:1.

Further Reading

Franzè, S., Cilurzo, F. and Minghetti, P. (2015) Insulin Biosimilars: The Impact on Rapid-Acting Analogue-Based Therapy BioDrugs 29:113-121.

Klint, J. K., Senff, S., Saez, N. J., Seshadri, R., Lau, H. Y., Bende, N. S., Undheim, E. A. B., Rash, L. D., Mobil, M. and King, G. F. (2013) Production of Recombinant Disulfide-Rich Venom Peptides for Structural and Functional Analysis via Expression in the Periplasm of E. coli PLoS One. 8(5):e63865.

U. S. Food and Drug Administration (2022) 100 Years of Insulin U. S. Food and Drug Administration. Accessed March 10, 2025.

About the Illustration

It is hard to write about insulin without giving a nod to Dorothy Crowfoot Hodgkin’s beautiful models. The illustration for this post is my drawing of one of her models, which resides at the Science Museum in London (object 1991-286/1). I have concentrated on the negative spaces between the copper, plastic and steel components to highlight the exquisite shapes. (Promarkers, felt-tip pens and pencil on paper, A3.)

Resources

Build a paper model of insulin with resources provided by the Protein Data Bank - Protein 101 Training and Outreach Portal.