Insulin purification: IEX, RP-HPLC and zinc crystallization

What is left after enzymatic conversion: impurities that look like insulin

The solution after enzymatic conversion is not all insulin.

It contains precursor that was not cut, molecules with arginine left on the end of the B chain (arginine-insulin and diarginine-insulin), des-Thr(B30) insulin, which lacks the terminal threonine, and deamidated forms in which some amino acids have changed. Host cell proteins and DNA are in the same solution too.

The trouble is the molecules that differ from the product by one amino acid or one charge. Des-Thr(B30) insulin appears repeatedly in the patent literature as an impurity that enzymatic conversion tends to produce and that is hard to separate. The challenge is a choice: suppress its formation by adjusting the conversion conditions, or separate it in purification.

You are purifying against molecules almost identical in appearance and properties, so everything depends on choosing the axis along which to separate them.

Why can't catalog binding capacity pick the resin?

Because the number is tied to the size and charge of the protein that was measured.

Resin makers measure with standard proteins such as bovine serum albumin or lysozyme and publish the result. Insulin has a molecular weight of about 5,800 and an isoelectric point (the pH at which the net charge is zero) of about 5.3. It differs from the standard proteins in size and in where its charge sits, so there is no guarantee of the same number. What to ask the maker is not how big the capacity is, but which protein the DBC was measured with, and at what pH and flow rate. A view for antibodies is in choosing an ion exchange resin.

A resin with large capacity is also not necessarily good at separating. Three kinds of resin are designed for different aims. High-capacity types such as Toyopearl GigaCap and POROS XS are built to capture a large reaction volume quickly. Resins with smaller particles for finishing, such as Capto ImpRes, separate close molecules. Tentacle types, such as Fractogel EMD, place the ligand at the end of a chain that extends from the backbone and can pick up molecules with close charges.

Do not rank candidate resins by catalog numbers. Compare them in small columns with your own post-conversion solution. What to look at is not binding capacity but how far apart the impurity next to the product comes out.

Binding capacity is a number for deciding whether a resin serves capture or finishing; it does not guarantee how easily insulin will separate.

Why use two steps on different axes, ion exchange then reversed phase?

Because no single axis separates every impurity.

The backbone of the purification is two steps that separate along different axes, placed in series.

The first step, ion exchange, separates by charge. Cation exchange binds the insulin to the resin, and raising the salt concentration gradually elutes molecules with slightly different charges one after another. A deamidated form gains negative charge because an asparagine becomes an acidic amino acid, so charge is the difference that can pick it up.

The second step, reversed phase, separates by hydrophobicity. The insulin binds to hydrophobic particles and is eluted by raising the fraction of an organic solvent such as acetonitrile. Molecules with the same charge but a slightly different hydrophobicity, because of shape or an amino acid difference, come apart here. How the acid you add changes the separation is explained in acid in the mobile phase for reversed-phase HPLC (in Japanese). A two-step purification with cation exchange followed by reversed phase has examples in the patent literature.

Which axis works for which impurity differs by substance and by condition. For separating des-Thr(B30) insulin, the patent literature has tried both ion exchange and reversed phase, but because the difference is a single amino acid, it states that no efficient way of separating had been found. Treat "moves by charge" and "moves by hydrophobicity" as a rule of thumb, and run your own list of impurities through both columns to confirm.

Particle size in reversed phase is a tug of war between load and resolution. Large particles raise the amount you can run per batch but reduce the ability to separate molecules close to the product. Small particles raise resolution at the price of smaller batches. Whether the risk of related substances exceeding the specification or the risk of smaller batches weighs more differs from process to process. Because the process uses large volumes of organic solvent, include explosion-proof equipment and solvent recovery in the cost comparison.

Design the steps so that each handles impurities along its own axis, charge or hydrophobicity.

Why does zinc crystallization count as the last purification step?

Because molecules that do not fit the crystal lattice have difficulty entering it.

Crystallization is a step that solidifies and takes out the product, and at the same time it is the last stage of purification.

Insulin holds a zinc ion at histidine 10 of the B chain, and three dimers assemble into a hexamer. Crystals are these hexamers arranged in a regular pattern, and molecules that do not fit the pattern have difficulty entering the lattice. Deformed molecules and impurities that stay dissolved are left in the mother liquor. Molecules very close to the product are thought to mix into the lattice more easily, so crystallization is not a replacement for chromatography. It is a finishing step placed after it.

Conditions are set along two axes: solubility and the speed of crystal formation. The isoelectric point of insulin is around pH 5.3, but crystallization is examined somewhat away from it. A study with human insulin (Cryst Growth Des, 2022) examined pH 6.0 to 6.7. It showed that raising the pH raised solubility about 5 times, while nucleation started up to 8.6 times faster. The insulin dissolves more easily and yet crystals form more easily, so pH is not a condition you can simply move in one direction. The type and amount of zinc salt are narrowed down along the same axes.

In laboratory studies, a common setup combines a reactive crystallizer with instruments that follow the number and size of crystals in situ (such as Mettler-Toledo's FBRM and PVM), and sets conditions by watching when to add seed crystals and how supersaturation (a state in which the solution is more concentrated than it can hold dissolved) changes. Washed crystals are dried gently so that particle size is not broken. The pharmacopoeial monograph has a specification for zinc content, so the amount you add and the amount the wash removes must be kept within that range.

Crystallization conditions are not the ones that solidify fastest; they are the ones that satisfy solubility, nucleation rate and the zinc content specification at the same time.

Where do insulin purification projects go wrong?

In four places, all of them a mix-up of a number or an order.

The first is choosing a resin by catalog binding capacity. A number measured with a standard protein does not carry over to insulin purification as it is.

The second is using a high-capacity resin in the finishing step as well. Resins built for capture hold a large volume quickly, but they are not suited to separating close molecules.

The third is fine-tuning purification conditions while the upstream process is still moving. If the precursor design, refolding or enzymatic conversion changes, the line-up of remaining impurities changes. Purification fixed too early has to be redone as soon as the upstream moves.

The fourth is treating crystallization as plain solidification. If you are satisfied that crystals formed without varying pH and zinc salt, room is left in both yield and purity. Do not expect crystallization to clear the related substances that chromatography could not remove, either.

Insulin drug substance is then made into a drug product. Rapid-acting insulin relies on how fast the hexamer comes apart, NPH on a suspension of protamine and zinc crystals, and long-acting glargine on the property that an acidic solution precipitates at the neutral pH under the skin (insulin analog design and manufacturing (in Japanese), choosing formulation pH and buffer (in Japanese)). The zinc content and crystal properties of the drug substance are the premise of formulation design.

Choose resins by role and with your own solution, settle the upstream process first, and treat crystallization as part of purification.

What to check before you build or audit a purification process

Look at these points when you put a purification process together, or when you ask a drug substance supplier about theirs.

1. For the resin's binding capacity, which protein was measured, and at what pH and flow rate? Is there a value for your own reaction solution?
2. Do you have a list of the related substances to remove, and have you confirmed for each whether it moves by charge or by hydrophobicity?
3. Are capture resins and finishing resins chosen separately by role?
4. Did you lean the reversed-phase particle size toward load or toward resolution, and why?
5. Did you look at the crystallization pH and zinc salt in terms of both solubility and the speed of crystal formation?
6. Is the zinc content likely to fall within the pharmacopoeial specification?
7. If the precursor, refolding or enzymatic conversion conditions change, is there a procedure for redoing the purification conditions?

Products can be compared on the directory pages for cation exchange resins, reversed-phase HPLC columns, crystallizers and drying and milling equipment.