Modeling stationary phases in chromatography: Particles, membranes, monoliths and fibers

Chromatographic separation fundamentally relies on the selection of suitable stationary phases. Conventional packed beds of porous particles have been supplemented by membranes and monoliths that show superior performance for certain molecule classes. In April 2020, Cytiva launched the Fibro line of adsorbers that represent another class of stationary phase. These functionalized cellulose fibers promise to combine high flow rates with high binding capacity.

This article discusses the pros and cons of the different materials and how they perform when it comes to mechanistic modeling. While traditional mechanistic modeling approaches have been derived for porous particles, we have shown that membranes, monolithic and polymeric fibers can be simulated accurately and easily.

Chromatography stationary phases and how to model them

Four different phases in a chromatographic system
Figure 1: The different phases in a chromatographic system: Interstitial volume of the mobile phase (a), liquid film around the beads (b), pore volume of the mobile phase (c) and stationary phase (d). Copyright © by GoSilico GmbH

Chromatography stationary phases come in a variety of forms and formats. The most common form is a packed bed of porous particles. As illustrated in Figure 1 (a), the molecules in the mobile phase are transported through the fluid outside of the particles, and then enter the particle’s pore system (Fig. 1b) and diffuse inside the pores (Fig. 1c).

The physical or chemical properties of the stationary phase are chosen such that some components of the injected sample are retained more strongly than others. In order to model the stationary phases of a chromatography column, the actual material properties pose severe restrictions on the applicability and suitability of a specific model.

Depending on the structure of the particles’ backbone, pore diffusion can limit mass transfer leading to visible tailing in the peaks of the chromatogram. If this effect is pronounced and important for example when using UV-based pooling criteria, it must be considered during model building.

The default case for mechanistic chromatography modeling: Porous particles

Different chromatography models exist with different level of details. ChromX includes several pore models with increasing complexity. The so-called Lumped Rate Model combines the film diffusion (Fig. 1b) and pore diffusion (Fig. 1c) and calculates average concentrations for the whole liquid in the pore (Fig. 1b). This concentration may differ from the concentration of the fluid outside of the particles (Fig. 1a). In our experience this rather low complexity model is well applicable to the more modern and rigid materials such as the Capto™ resins by Cytiva and POROS™ by Thermo Scientific.

More complex models such as the General Rate Model with and without surface diffusion is usually needed for resins based on Sepharose for example.

Other stationary phases have no requirement to take pore and surface diffusion into account because micro pores are simply non-existent.

Modeling complex geometries: Membrane and monolithic adsorbers

Membrane adsorbers for example only have large pore sizes and have thus been identified as promising stationary phases for the processing of bionanoparticles. Already in 2014, a capture step for a virus-like particle (VLP) by BioNTech on a membrane adsorber by Sartorius have been simulated with ChromX. Another publication followed recently.

Depiciton of a spiral wound membrane adsorber Copyright by GoSilico
Figure 2: Spiral wound membrane adsorber. Copyright © by GoSilico GmbH

Membrane adsorbers

The geometries of membrane adsorbers are usually either stacked flat sheet or spiral wound modules with axial or radial flow, which can all be simulated with ChromX. Regarding separation performance parameters such as resolution, peak width, and dynamic binding capacity, there are no outstanding differences between axial and radial flow geometries.

However, radial flow geometries are especially suited for large-scale applications by virtue of lower pressure drops: Wide, short axial flow columns can be similarly replaced with low-pressure narrow, tall radial-flow columns without a significant change in performance. The combination of both radial flow geometry and convection driven mass transport in membrane adsorbers allows high throughputs and productivities for biopharmaceutical products.

Monolithic adsorbers

The same advantages apply to monolithic adsorbers shaped as a cylinder ring. In comparative studies for ion-exchangers, we observed even better selectivity on monoliths by BIA Separation compared to membranes. An example of a VLP capture step on a monolith can be found in this recent case study.

In comparison to packed-bed columns with frits, the dead volume in capsules containing membrane or monoliths are typically significantly higher and introduce mixing effects that need to be taken into account. In ChromX, this can be done by adding virtual stirred-tanks and plug-flow reactors.

Mechanistic modeling for fibers-based media

The macroscopically observable behavior of membranes and monoliths is very similar to a new type of stationary phases: fibers. They can be prepared from various base polymers, including natural and synthetic ones, and are available in different formats, i.e. different shapes, lengths, and structures. Recent advances in the fabrication of high surface area fibers and in surface modification protocols have made it possible to prepare fiber-based adsorbents with high capacities and low pressure drops at high flow rates.

In 2016 we assessed whether mechanistic modeling can be performed on fiber-based adsorbents. With a column randomly filled with short cut hydrogel grafted anion exchange fibers, we tested whether tracer, linear gradient elution, and breakthrough data could be reproduced by mechanistic models. Successful modeling was achieved for all of the considered experiments, for both non-retained and retained molecules. The model accurately accounted for the convection and dispersion of non-retained tracers, and the breakthrough and elution behaviors of three different proteins with sizes ranging from 6 to 160 kDa.

3 chromatograms comparing experimental and simulated breakthrough experiments on the fiber column with insulin, BSA, and GO
Figure 3: Comparison of experimental (dashed lines) and simulated (solid lines) chromatograms for breakthrough experiments on the fiber column with (A) insulin, (B) BSA, and (C) GO. Dotted lines represent the simulated salt profiles at the column outlet. Flow rate: 1 ml/min. Protein concentrations: 2 g/L.

Conclusion and outlook

When it comes to working with particle-based resins, the good news is that modeling has become easier over the last decades. The modern rigid particles do not show a strong pore diffusion limitation anymore and consequently less tailing of peaks in the chromatograms. That makes the life of the modeler easier who does not have to care about this phenomenon anymore.

The model complexity for membranes, monoliths and fibers is even lower as no micro-pores must be considered. However, we have to differentiate between axial and radial flow as the geometry has an impact on the resulting peak shapes. Both operation modes are readily available in ChromX. Special attention must be paid to dead volumes in the adsorber capsules that might introduce mixing effects. With the promising results we obtained for fibers in 2016, we are confident that we can work equally well with the new Fibro line of adsorbers and looking forward to first data coming in.

Whatever floats your boat – or purifies your proteins – ChromX is prepared.