Ion-exchange chromatography has evolved – and so has our understanding

Tuning up load densities, highly engineering molecules, using state-of-the-art ion exchangers – these key enablers for efficient processes pose new challenges in form of unusual peak shapes and irregular behavior, demanding a new interpretation of adsorption. GoSilico addresses these needs by having introduced the Colloidal Particle Adsorption (CPA) model, an adsorption isotherm based on a colloidal description of proteins.

The best team wins: collaboration of industry partners to resolve the mechanism behind inexplicable peaks

A central issue found in industrial chromatography are repeatedly occurring peak shoulders and “shark fin” peaks at high load densities that complicate experimental and in-silico process development. The lack of process understanding when it comes to the mechanism behind these effects hinders the investigation of possible root causes and decelerates the correct tuning of process parameters.

What got apparent is, that uncommon peak shapes are not restricted to one molecule, resin or process. Thus, taking up the challenge of finding an explanation for peak shoulders and shark fin peaks requires cooperation. For the first time, the who-is-who of German biopharma

  • Bayer
  • Boehringer Ingelheim
  • Rentschler
  • Roche

combined forces by sharing data from their antibody polishing steps to accomplish this scientific mission and GoSilico took, almost literally, the burden off their shoulders.

Resins, modes, processes: new challenges to prove CPA’s versatility

The scope of available data covered three different cation exchange resins, as shown in the table below. All adsorbers are functionalized with strong cation exchange ligands. In total, several bind and elute experiments with six different mAbs were performed, including salt elution at different pH values, as well as pH-induced elution at different salt concentrations. Moreover, the load densities were varied. An overview of all experiments can be found in the recent publication of Briskot [1].

Table 1: Overview of the resins utilized in the experiments simulated with the CPA model [1]. All adsorbers are functionalized with strong cation exchange ligands with a pK value of pH 2.3.

Capto S ImpActCytivaHighly cross-linked agarose
Poros 50 HSThermo Fisher ScientificPoly styrene divinyl benzene-based polymer beads
Fractogel EMD SO3-Merck MilliporeSynthetic methacrylate-based polymer beads with tentacle-like structures grafted onto bead surface

Salt, pH, overloading: one model to fit them all

Figure 1 shows the results of salt and pH gradient elution experiments for all three resins, together with the fitted curve by GoSilico’s CPA model. The remarkable results speak for themselves – CPA was able to describe all peaks adequately, independent of the respective resin, mAb or type of elution.

Figure 1: From left to right: Salt gradient experiments of mAb1 on Poros 50 HS according to [2], mAb2 on Fractogel EMD SO3- according to [4], mAb3 on Capto S ImpAct according to [1], and a pH gradient experiment with mAb6 on Poros HS according to [1]. Blue lines represent experimental data, orange lines indicate simulated data based on the CPA model. The ionic strength or conductivity of the mobile phase is shown in grey, the pH value is displayed in dark green.

For the first time, high-ranking industrial partners – Bayer, Boehringer Ingelheim, Rentschler, and Roche [1] – have combined data from their antibody polishing steps. It is precisely the variety of molecules and materials that allows us to understand what otherwise remains hidden in the column. CPA was able to reproduce all of them – a true all-rounder based on one consistent approach.

Revolutionized and consistent: One explanation providing in-depth process understanding

While a well-fitted peak is a good starting point for in silico process development, the superordinate goal of mechanistic models is to provide explanations and understanding of the occurring effects. CPA considers the available free surface area as the limitation for binding. Based on this understanding of the CPA model, the following explanation was proposed for the observed peak shoulders.

Figure 2: Schematic representation of the column after loading (top) and during elution (middle) at low (left) and high (right) load capacities. White areas indicate unoccupied, colored areas indicate occupied areas, respectively. A schematic representation of the resulting chromatogram is depicted (bottom), grey lines from light to dark grey represent increasing load densities.

Figure 2 compares a column loaded with a load density around half its capacity (left) and an overloaded column (right). The distribution of bound protein along the column during the process was simulated. It could be shown that for the half-loaded column right after loading (Fig. 2, top left), the upper part of the column is highly saturated. Towards the bottom part of the column, binding sites are unoccupied, providing free surface area for binding. With the initiation of elution (Fig. 2, middle left), the increase in ionic strength leads to protein desorption at the top of the column. Once these proteins reach the unsaturated bottom part of the column, they abruptly decelerate due to re-adsorption onto the unoccupied areas and still favorable binding conditions. Hence, the elution front is thwarted in the bottom part of the column. At the same time, protein steadily arrives from the top of the column, leading to a compression and concentration of the elution front, which results in a very steep peak shape.

Briskot [1] could show that after exceeding a load density higher than approximately half the column capacity (Fig. 2, left), the position and height of the peak maximum remains unchanged – an indicator that the elution front cannot be more concentrated.

If now the load density is further increased, in other words, an overloaded column is regarded (Fig. 2, top right), the unsaturated part at the bottom of the column is very small or non-existent – all binding sites are occupied. Consequently, as soon as the elution starts (Fig. 2, middle right) the velocity of the elution front is almost comparable to a non-retained species due to the highly occupied area and steric hindrance in the column. Moreover, the concentration front has less time or distance to be concentrated, as barely any re-adsorption takes place. This leads to an immediate and rapid elution and more dispersed frontal part of the peak (Fig. 2, bottom right). Consequently, the formation of the observed shark fins and shoulders for high load densities takes place.

Practical applicability: Keeping the calibrational effort low

CPA provides one consistent explanation while showing outstanding simulation results for different mAbs, resins and load densities. Despite its superb performance, it does not extend the calibration complexity. On the contrary: CPA’s model parameters describe the biomolecule and not the process – this means that once a biomolecule is characterized, the transfer to another process or resin is significantly simplified. A perfect complement to speed up your calibration is to use pre-characterized columns – even when not using model-based process development, the information about column and resin lot provides valuable insights.

Stop guessing. Go colloidal.

Extraordinary and versatile modeling performance coupled with in-depth process understanding – the CPA model paves the way for the future of productiveness and robustness at high column loadings.


[1] T. Briskot, T. Hahn, T. Huuk, G. Wang, S. Kluters, J. Studts, F. Wittkopp, J. Winderl, P. Schwan, I. Hagemann, K. Kaiser, A. Trapp, S. M. Stamm, J. Koehn, G. Malmquist, J. Hubbuch, Analysis of complex protein elution behavior in preparative ion exchange processes using a colloidal particle adsorption model, Journal of Chromatography A 1654 (2021) 462439.

[2] T. C. Huuk, T. Hahn, K. Doninger, J. Griesbach, S. Hepbildikler, J. Hubbuch, Modeling of complex antibody elution behavior under high protein load densities in ion exchange chromatography using an asymmetric activity coefficient, Biotechnology Journal 12 (3). doi:10.1002/biot.201600336.

[3] J. M. Mollerup, A review of the thermodynamics of protein association to ligands, protein adsorption, and adsorption isotherms, Chemical Engineering and Technology 31 (6) (2008) 864–874. doi:10.1002/ceat.200800082.

[4] J. Diedrich, W. Heymann, S. Leweke, S. Hunt, R. Todd, C. Kunert, W. Johnson, E. von Lieres, Multi-state steric mass action model and case study on complex high loading behavior of mAb on ion exchange tentacle resin, Journal of Chromatography A 1525 (2017) 60–70. doi:10.1016/j.chroma.2017.09.039.