OR WAIT 15 SECS
Volume 30, Issue 7
The authors present a shift toward more integrated purification processes.
Today the purification processes for monoclonal antibodies (mAbs) are rather well defined, including a first capture based on protein A, followed by a viral inactivation with low-pH treatment and usually two polishing steps based on ion exchange or other mode of interactions. So far, all these steps are performed in a discontinuous, batch manner (1,2). On the development side, the main strategy is often to optimize each step individually and bring each of them together, assuming it would produce the best process.
For some years now, the concept of continuous and intensified manufacturing has been gaining interest and evaluated by many therapeutics producers. The nature of “continuous processing” has also been discussed and it is not always clear what it actually means, and if what one expects to gain is really there to be found (3). Nevertheless, process intensification is quite well understood as the way to make processes smaller, smarter, and more efficient, whether or not it is continuous and to what extent.
This paper presents the COMPAC2T platform (Merck KGaA), a process for intensifying the purification train with a holistic, more integrated approach (see Figure 1) and the latest advances on the chromatography train.
Figure 1. Concept of the COMPAC2T purification platform. all figures are courtesy of the authors.
Protein A remains the most straightforward purification step to be placed at the entrance of the purification train. It has proven to be highly efficient and reliable, easy to implement at any scale and backed-up by significant knowledge in industry, plus it benefits from a clear support from health authorities. Protein A manufacturers have been able to propose better resins with higher capacities and modified ligands to allow elution at higher pH or NaOH to be used as a regeneration/sanitization agent. The main challenge today with Protein A is to find the right processes to allow a higher productivity in order to cope with the evolution of the upstream production train. In the past decades, titers have dramatically increased in the bioreactor to reach typical values of 5-7 g/L, if not more. The most common strategy has been to increase column size to increase the cyclic productivity of the batch process: a simple, successful strategy but a rather non-efficient one regarding the productivity, materials and space utilization, and costs.
In the presented platform, a multi-column chromatography process is evaluated as an alternative to the traditional batch-capture process. This process is based on the scheduling of several small chromatography columns allowing saturation of the resin while increasing flow rates. The bed height is varied during the loading step by sequentially adding or removing columns from the loading zone. This process scheduling allows the resin to be saturated without losing any mAb by breakthrough. Saturated columns are then submitted to the other steps (i.e., washing, elution, regeneration, and equilibration) in a batch-wise manner while loading is continued on other columns. In the present case, the multicolumn system used is the BioSC (Novasep, Pompey, France) together with simulation-optimization software (BioSC Predict, Novasep). Detailed explanations of the multi-column process used can be found in Ng et al. (4) and Girard et al. (5). The choice to evaluate and implement multi-column processes results from the following facts:
Multi-column chromatography technologies should be evaluated for each manufacturing strategy and objective because the different constraints associated could significantly modify the process set-up. To better define how the multi-column technology would look like for a capture step associated with fed-batch bioreactors, the authors evaluated the impact of several Protein A resins on the design of the process. The objective was to increase the production rate (kg/h) while decreasing the resin volume and contain the system in pre-existing production suites. A more detailed analysis based on similar objectives is available in Hilbold et al. (6).
Table I summarizes the production rate for different resins for different harvest concentrations, together with the number of columns and the resin volume considered.
Table I. Comparison of several Protein A resins in batch and multi-column mode for harvests at 1 and 10. mAb is monoclonal antibody.
Resin A Monocolumn
Resin A BioSC
Resin B BioSC
mAb concentration (g/L)
Number of column
Total packed bed volume (L)
What is clear based on these data is that there is a direct gain by switching from the mono-column process to the muli-column process for the resin A. The production rate is at least doubled, if not improved seven-fold. This improvement is mainly due to the fact that non-loading steps are hidden behind a semi-continuous loading of the harvest while a higher saturation of the resin can be reached.
The data also show that all resins are not impacting the multi-column process in the same way: their respective behavior regarding pressure drop, total capacity, and capture kinetics will set the limitations of the process. The number of columns is increasing with the harvest concentration since by increasing concentration of the harvest, the loading phase becomes shorter compared to the other steps and an additional column is required more quickly than at low mAb concentration. As a result, the time available to perform non-loading steps becomes smaller, until an additional column is required to maintain the process performance. For all resins, there is a rather similar resin volume requirement compared to the batch. On the other hand, the production rates are systematically higher, demonstrating that resins capacity is better used. Resin C is mainly impaired by a narrower pressure drop window limiting the injection velocity, while remaining better than the mono-column process at high concentration. Resin A has a favorable capture kinetic compared to Resin B: its breakthrough profile is less impacted by reducing the residence times and remains steeper. A systematic mapping of the performance of the multi-column process should be performed to better understand when it is relevant and applicable.
So far, the multicolumn system process has been tested on several mAbs, demonstrating robust results on both the yield and the quality sides. More than 70 cycles have been performed on a three-column process without significant performance losses.
Several questions have still to be answered. For example, the impact of high saturation of the resin at each cycle remains uncharacterized (or unclear). While this issue mostly concerns the column lifetime, variations are expected on all steps (i.e., washes, elution, and regeneration). Also, the multi-column approach presents questions on how the manufacturing strategy could be adapted to be more efficient. For example, instead of using a large column for three to four cycles per campaign, would it be better to use small columns for 20 cycles in the same time windows? The resulting investment would be much lower and the management of the resin storage simplified. One can also consider using pre-packed columns for three to four campaigns and discard them after the validated lifetime is reached.
The polishing steps are often performed using a combination of chromatography steps, typically two in addition to the Protein A capture, often based on ionic and/or hydrophobic interactions. Most of the time, the first polishing step is performed in bind-and-elute mode, while the last one is performed in flowthrough (FT) mode. The process aims to execute both polishing steps in flowthrough mode and allow a direct connection between the two steps. In such a setup, the output of FT1 becomes directly the input of FT2, thanks to an intermediate automatic, online pH/conductivity adjustment.
The major advantage of FT mode is well known: the objective is to capture the impurities that are much less present in proportion compared to the protein of interest. By focusing more on the capture of impurities, the capacity of antibody that can be processed in one cycle is significantly extended, typically a four-fold increase, depending on the impurity load. The challenge here is to actually find conditions allowing the capture of a large variety of impurities including host cell proteins (HCPs), mAb aggregates, and clipped mAbs, without compromising the yield in monomeric form. A first introduction of the FT-FT platform can be found in Xenopoulos et al. (7).
By connecting directly FT1 and FT2, downtimes and intermediate storage are avoided, thus improving the productivity of the polishing platform. This approach requires the development of a robust online, automatic pH/conductivity adjustment system. This approach will require the definition of a preliminary buffer that will remain quite similar in composition from FT1 to FT2 and should be compatible with both steps. The sizing of each FT step is also important to keep capacities and residence time in line from one step to another.
Figure 2 shows the cycling of a FT-FT platform over 50 cycles at lab scale (5 mL columns). The yield was good with a consistency of approximately 80%. Aggregates level (around 2-2.5%) and HCP level (below 10 ppm) were within specifications, with starting values of, respectively, 5% and 3500 ppm. The variations of aggregates level and increases of yield starting at run 22 are due to a readjustment of the loading pH. Overall, no pressure drop increase has been observed and pH/conductivity profiles remained stable.
Figure 2. Evolution of yield and quality attributes for a monoclonal antibody (mAb) A over 50 cycles on the flowthrough FT-FT platform. HCP is host cell protein.
A proof of concept combining a Protein A capture in multicolumn system mode with a FT-FT platform as described previously demonstrated the following results. The multi-column process involved three columns of 5 mL loaded at a capacity of 62 g/L, with an elution every 26 minutes. The FT-FT platform used two columns of 5 mL loaded up to 400 g/L, consequently several Protein A elutions had to be pooled to performed one FT-FT cycle. This intermediate storage facilitates the implementation of a classical low pH inactivation, without impacting the process performance. The process was run on 20 cycles of Protein A with five cycles of polishing platform, demonstrating a good stability and a global yield of 78%. With such an approach, estimations show that the productivity could be improved up to 25 gmAb.h-1.Lresin-1, versus 3 gmAb.h-1.Lresin-1 for the traditional process including a monocolumn Protein A step associated with a bind-and-elute step, an intermediate storage and a flowthrough step.
The process developed also implies a higher degree of automation, a better integration of the steps, and the introduction of relevant analytical tools. Cycle times tend to become smaller with a more pronounced interdependence between all the steps, implying a smaller tolerance to mistakes and downtimes to keep the process running. Therefore, decisions should be taken quickly to readjust the process parameters if needed to ensure a consistent product quality. To do so, process analytical tools need to be implemented to follow the relevant quality indicators, with an important role of the data recovery strategy and more sophisticated process control approaches based on holistic models. Most of the existing processes require between 10 and 15 buffers to be prepared in advance, implying an important storage capacity based on stainless steel tanks. The concept also aims to implement an on-demand delivery buffer system to diminish the need of such large capacities. The principle is to simplify the standard buffer platform and prepare specific mixes only when required to avoid the need for large capacities and the management of these infrastructures.
The platform presented by the authors is a first step toward better integrated and intensified purification processes, as well as a better understanding of purification processes, based less on the gathering of individually optimized steps and more focused on the definition of a process globally more efficient and robust.
1. J. H. Vogel, et al., Biotechnol. Bioeng. 9, 3049-3058 (2012).
2. A. Abhinav and T. Jörg, Trend Biotechnol. 28 (5), 253-261 (2010).
3. K. Konstantinov and C. Cooney, J Pharm Sci 104: 813-820 (2014).
4. C. K. S. Ng et al., Food and Bioproducts Processing, 92(2), 233-241 (2014).
5. V. Girard, et al., “Large-scale monoclonal antibody purification by continuous chromatography, from process design to scale-up,” Journal of Biotechnology, Vol. 213, 65-73 (2015).
6. N.-J. Hilbold, et al., “Evaluation of Several Protein A Resins for Application to Multicolumn Chromatography for the Rapid Purification of Fed-Batch Bioreactors,” Biotechnology Progress, In press, 1-13, (2017).
7. A. Xenopoulos, Journal of Biotechnology, 1-12 (2013).
Volume 30, Number 7
When referring to this article, please cite it as -J. Hilbold et al., “Intensification of a Chromatography Platform," BioPharm International 30 (7) 2017.