BIOPROCESS TECHNICAL

Making Downstream Processing Continuous and Robust A Virtual Roundtable S. Anne Montgomery, Cheryl Scott, and Peter Satzer, with Margit Holzer, Miriam Monge, Ralph Daumke, and Alexander Faude

urrent biomanufacturing is driven to pursue continuous processing for cost reduction and increased productivity, especiallyC for monoclonal antibody (MAb) production and manufacturing. Although many technologies are now available and have been implemented in biodevelopment, implementation for large-scale production is still in its infancy. In a lively roundtable discussion at the BPI West conference in Santa Clara, CA (11 March 2019), participants touched on a number of important issues still to be resolved and technologies that are still in need of implementation at large scale. Below, moderator Peter Satzer (senior scientist RENTSCHLER BIOPHARMA (WWW.RENTSCHLER-BIOPHARMA.COM) with the Austrian Center of Industrial Biotechnology, Vienna) summarizes key points raised in that session. Based on Requirements for integration without the need for hold his highlights, BPI asked a number of Continuous Downstream tanks between unit operations requires industry representatives to comment on further exploration. those points further, and their Applications Chromatography Operations: Some responses follow. by Peter Satzer unit operations such as flow-through chromatography and single-pass, Minimizing Buffer and Hold Tanks: One tangential-flow filtration (SPTFF) can be critical parameter is integration of unit integrated easily into a continuous Product Focus: Biologics operations on manufacturing shop downstream process scheme at any floors using minimum numbers of point because they offer constant inflow Process Focus: Downstream buffer tanks and hold tanks. The benefit and outflow of material. Such operations processing of continuous manufacturing can be can be called truly or fully continuous. realized fully only if auxiliary Other unit operations — especially Audience: Process developers, equipment for buffer preparation and bind–elute chromatography — are much manufacturing, operations hold tanks are minimized between harder to implement because they don’t Keywords: Filtration, Buffer downstream operations. Although provide truly continuous operation, but preparation, chromatography, continuous buffer preparation and rather produce individual elution peaks virus safety reduction of buffer hold tanks are being with fluctuating product and impurity investigated (and also were the subjects concentrations. Usually the unit Level: Intermediate of BPI West presentations), direct operation following such a step cannot

38 BioProcess International 17(6) June 2019

17-6-Satzer-RT.indd 38 8/28/2019 10:29:21 AM Roundtable Participants for batch definition and to minimize drug-substance manufacturing. A couple Peter Satzer (senior scientist with the product loss in case of contamination, a decades of production improvements have Austrian Center of Industrial Biotechnology, broad residence-time distribution created challenging process streams for Vienna) smooths process irregularities and leads separation and purification. Can continuous Margit Holzer principal at Ulysse Consult to a more robust process. The trade-off downstream processes offer a partial between those two approaches will need solution to this problem? BPI editorial advisor and Miriam Monge to be discussed in the future. Satzer: Continuous downstream head of the MAb and intensified PCC Chromatography: PCC can be processes can offer solutions in two bioprocessing market segment at Sartorius detrimental to our understanding of separate ways. The first way is to Stedim Biotech SA (whose comments residence-time distribution because it increase downstream equipment use by include insights from colleagues Ganesh Kumar, Tom Erdenberger, and Hemanth can split it during an operation. A PCC using equipment 24–7 and therefore Kaligotla, members of the Sartorius MAb operation will produce one or two increasing downstream productivity. and intensified market segment team) residence-time distribution peaks The second way is through a change in depending on the state of the operation, mindset, in which up- and downstream Ralph Daumke biotechnology market and no easy model can be applied. If operations are integrated directly rather manager life science, FILTROX Group bind–elute polishing also is than remaining distinct and separate — Alexander Faude, PhD, director DSP implemented using PCC, the resulting which forces them into direct development at Rentschler Biopharma residence-time distribution can be split communication. That opens the (whose comments include those of into up to four individual distributions possibility of solving problems earlier colleagues director DSP process Thilo Grob traveling through the system, which (upstream), where solutions might be design/validation and Anja Trapp senior complicates decisions about when to easier to achieve. scientist) discard material. Currently, this issue is Holzer: Great improvements have too poorly understood and modeled. It been made in cell-specific productivity, handle such changes in flow and either is not addressed, is addressed media development, and process design concentration directly, requiring with wide safety margins for discarding during recent decades, resulting in some inclusion of surge vessels and hold material, or is addressed by cases up to 100× increased product tanks that interrupt the process. This implementing a hybrid process that uses concentration in fermentation broth. makes continuous periodic batch manufacturing after a PCC unit The product output of the same countercurrent chromatography (PCC) a operation. upstream installation could be unit operation that is harder to Alternative technologies to bind– increased proportionally. implement than other operations such elute chromatography such as flow- However, improvements in as flow-through chromatography and through chromatography, precipitation, chromatography media capacity and new, alternative purification methods, crystallization, and flocculation can membrane performance have not given including continuous precipitation. offer solutions to those issues in the the same range of productivity increase, Viral Inactivation: Current future to preventing integration instead doubling or tripling it. implementation of continuous problems. Developers of future Therefore, work is needed on process processing in the MAb world could make technologies should aim for a constant design to overcome the downstream use of viral inactivation steps as surge mass flow through all unit operations, bottleneck. Related design studies easily vessels. Biomanufacturers could fill enabling integration and a seamless show the great potential of continuous parallel vessels periodically and empty downstream process train. processing. them to provide constant flow and Monge: Recent advances in product concentration for subsequent upstream processing have led to unit operations. This might not be The Roundtable Discussion processes in which expression titers possible for all antibody and Based on the discussion highlights have increased beyond 10 g/L or even nonantibody products, however. above, BPI invited further insights from 20 g/L. Availability of concentrated fed- Alternative purification techniques such several advisors, end users, and batch options on a pilot and as continuous precipitation offer a supplier companies. Included below are commercial scale has resulted in DSP constant inflow and outflow of material additional comments from Satzer along handling product amounts ranging and prevent integration issues. with responses from participants listed from 5 kg to 40 kg per batch. On one Residence-Time: Another important in the box at the left. BPI invites further hand, process intensification efforts point to consider when thinking about comments and welcomes submitted have reduced overall footprint and time continuous integrated unit operations is manuscripts detailing how your in the upstream train by eliminating the residence-time distribution through organization approaches these issues. intermediary seeding steps. But this all unit operations. Understanding of has led to an increased footprint in this concept is limited, but it is critical Capacity Mismatch downstream processing (DSP) that has to ensure batch definition and ensuring We’ve heard recent discussion about a required some biopharmaceutical process robustness. Although a narrow “capacity mismatch” between upstream and companies to switch from single-use to residence-time distribution is preferred downstream aspects of biopharmaceutical hybrid/stainless steel technologies.

40 BioProcess International 17(6) June 2019

17-6-Satzer-RT.indd 40 8/28/2019 10:29:21 AM One way to address the challenge is Satzer: The final goal has to be considerable decrease in downstream to implement multicolumn preparation of buffers and media based buffer demand. Choosing the right chromatography at the capture step to on powders or concentrates. For buffer-management strategy would ease the burden or to have continuous concentrated buffers (and reduction of require a company not only to evaluate DSP with perfusion-based upstream different buffers used in the complete the buffer demand on a unit operation processing (USP) — enabling downstream stream), efforts have been and process level, but also to evaluate biopharmaceutical companies to made already and the first prototypes the impact of each proposed preparation debottleneck this “capacity mismatch” created (some presented at BPI West). and distribution concepts on mobility, because the amount of product to be Media tanks especially occupy a adjacency, and room classification. processed is distributed daily. The significant portion of a shop floor and As far as continuous processing is approach also provides high flexibility either take up significant space or have concerned, choosing ready-made in product output, reduces the overall to be refilled regularly (requiring buffers connected at the point-of-use, DSP footprint, and enables closed personnel, analytics, maybe cleaning, along with in-line dilution and stream processing and cleanroom and so on). Buffer preparation is an conditioning for steps with low buffer- declassification. With the same scale for essential part of DSP, so development of demand could be an option. So could clinical and commercial production, the continuous buffer preparation from choosing a buffer-on-demand system technology transfer process is seamless. powders is fundamental to a fully with powder/concentrate that Daumke: Different options can help continuous process with minimal shop- eventually will be formulated into solve the mismatch. One option would space requirements. buffers and released in real time for be a single-use centrifuge, and another Holzer: Perfusion or chemostat cell steps with moderate to high buffer would be filtration. In the latter case, culture processing typically need more demand. alluvial filtration can be developed and media while significantly improving Adapting that approach would mean built as a continuous skid working plant productivity. Depending on the that buffer needs/demands in DSP could similarly to a chromatography system. process design of unit operations in DSP, be addressed continuously by a When capacity is reached in the first the amount of necessary buffers and modular, intelligent buffer skid that is filter, the system switches automatically solutions may be reducible. adjacent (e.g., in close vicinity) to the to the next filter, and so on. For example, continuous downstream DSP skid but in a different room. That Faude: Continuous processing processing (cDSP) could be achieved by would eliminate the need for high- contributes to facing DSP challenges in connecting staggered batch operations, volume single-use mixers for buffer two ways. First, it usually goes in hand but that would not reduce buffer preparation and significantly reduce the with increased automation compared consumption. Yet in the case of area required for buffer preparation and with conventional batch processing. For countercurrent multicolumn distribution. example, using more sophisticated chromatography processes or single- Faude: Buffer reduction in chromatographic devices allows pass TFF, typically more product can be continuous downstream processing manufacturing to run efficiently over 24 processed with less buffer. It is seems unable to adjust to increased hours. Second, the continuous processing important to analyze an individual media amounts used in USP. Using mode enables higher resin use, resulting product, plant, and strategy to come up buffer concentrates is a very interesting in lower buffer amounts needed for with an adapted, balanced, and cost- possibility for reducing downstream processing. Both can have significant effective processing platform. buffer volume substantially, especially impact on costs. Moreover, continuous Cost for buffer preparation, analysis, when DSP occurs in facilities that are processing can trigger optimization of and storage are significant for recovery not designed for continuous processing. simple process steps such as flow- and purification steps and can become through applications and should push an operational bottleneck for some Filtration forward the development of resins, plants. That might be not seen during Continuous filtration can involve problems membranes, and other technologies that clinical phases, but it becomes evident related to low flux. Does this — or do other can deliver fast mass transfer. during commercial production. Cost challenges — limit its potential in studies and risk assessments should downstream processing? What kinds of Buffer Preparation support decision making. Results often technological solutions are needed? Whereas continuous upstream processing bring about interest in working with Faude: Prepared backup filters placed uses more buffers and media, continuous concentrated solutions using in-line in parallel might solve the problem. downstream processing should reduce the dilution systems. Currently, Switching to a parallel filter could be number of buffers required for separation technologies that allow buffer or media automatic depending on the flux course. and purification operations. Does the end preparations starting from powders are Doing so might make filtration a simpler result balance out? And does it make more under development. and better-controllable process step sense to buy and store premade, ready-to- Monge: Performing chromatography compared with continuous column use buffers for continuous processing — or in multicycle, sequential batches with chromatography. to make them on demand based on smaller columns/adsorbers that require Satzer: Filtration can be made powders and/or concentrates? a high degree of saturation has led to a continuous in two different ways: by

42 BioProcess International 17(6) June 2019

17-6-Satzer-RT.indd 42 8/28/2019 10:29:21 AM exchanging filters or by using models and software could help us flow. So I believe that those technologies membranes in TFF mode (either single further characterize this unit operation, (along with flow-through pass or not) with regular cleaning and capitalize on historical data, and assist chromatography for polishing) will solve regeneration or. In the end, all filters in process design and control. Use of the issues of bind–elute chromatography. tend to foul over time in a continuous process analytical technologies (PAT) to That said, I think that bind–elute process and will cease to function at reveal rapid information about filter chromatography is here to stay. There some point. The exchange of filters and integrity or the amount of filtered might be separation and purification continuous filtration can be used (and product are of utmost importance as well. tasks for which no suitable precipitation must be used in the case of dead-end Monge: Continuous processing technology exists, and the filtration) and has been demonstrated created an additional driver to biopharmaceutical world is conservative by our group and others for viral understand further the technical and when integrating new methods. It will filtration. In my opinion, the technology operational aspects of establishing tend to implement more familiar exists and can be used, but large-scale filtration steps toward manufacturing methods (such as PCC), at least in the implementation is still missing. implementation. The worst-case criteria near future. Holzer: Many different filtration for batch-mode filtration do not Holzer: Control of residence time and principles (e.g., micro-, ultra-, and nano- translate to newer process scenarios. loading capacity is important for batch depth filtration and membrane Besides a lower volumetric flow and and continuous (countercurrent adsorption) are applied during prolonged exposure, significant multicolumn chromatography) bind– downstream processing. Most filters are challenges that need addressing include elute as well as for flow-through used in a frontal-filtration mode batch filter installation, integrity testing, and chromatography (where impurities are process, in which concentrations change the impact of process variations and bound and can break through if over time and reach the maximal limit physicochemical feedstock variations. residence times and process conditions of filtration capacity. Bioburden The challenges mostly involve virus are not controlled). Such applications reduction, depth, or sterile filters are filtration because the number of other require tighter control and more process examples. filtration steps is decreased in an understanding for continuous bind– Redesign of those filters and unit integrated processing approach. Small, elute chromatography because operations into a tangential continuous smart surge tanks are one way to reduce robustness is reduced. process mode often is not of any process many concerns. They can serve to bridge Flocculation and precipitation steps and economical interest. Therefore, the downtime and compensate for process for impurities and target products are most efficient way is to work with perturbations from other unit applied in several industrial production staggered-batch or parallel-batch operations. processes (e.g., impurity precipitation/ configurations. The flux is proportional Daumke: In my opinion, the whole flocculation mainly for products to the membrane surface area or the liquid flow needs to be optimized to expressed in microorganisms and number of filters used. reach the flow needed. Clarification the product precipitation in Cohn plasma TFF processes allowing for use of (for example) depth or alluvial fractionation). Based on the continuously run unit operations are filtration requires a certain flow to performance of protein A capture steps applied typically to cell separation, become effective. Therefore, other for antibodies, it might be difficult in buffer exchange, and product options need to be investigated (e.g., this case to compete with bind–elute concentration. Because the flow of a continuous centrifuges or TFF). chromatography. Such alternative continuously run operation usually is technologies need to be evaluated case lower than in a batch operation, Chromatography by case, taking into account technology is available or can be Continuous bind–elute chromatography precipitation behavior, solubilization designed properly (for example, in the involves concerns over residence time and conditions, processing times, product case of single-pass TFF). loading capacities. Do you see this as a quality, yields, and so on. Reducing the number of unit chance for more flow-through operations to Monge: Bind–elute chromatographic operations during downstream succeed? What about alternative operations are by nature discontinuous processing — such as by buffer purification technologies such as processes and can become a bottleneck. exchanges or concentration between precipitation/flocculation? Other techniques such as multicolumn steps — is a great help for reducing cost Satzer: Bind–elute chromatography and simulated moving-bed and improving productivity. This is cannot be run in fully continuous mode. chromatography can bring resin-based facilitated by the availability of Flow-through operations would be systems closer to a continuous mode. chromatography media that can preferred, but in many cases they cannot However, the approach requires accommodate higher salt concentrations offer the same purification efficiency as sophisticated hardware systems with and flow rates with more targeted buffer bind–elute chromatography. We tedious automation and validation choices. demonstrated that precipitation/ needs. The integration of viral filtration steps flocculation can yield similar purities to On the other hand, membrane-based into continuously run DSP lines affinity chromatography in bind–elute flowthrough polishing provides a cost- continues to be challenging. Process mode while offering a constant mass effective and highly productive

44 BioProcess International 17(6) June 2019

17-6-Satzer-RT.indd 44 8/28/2019 10:29:22 AM alternative to resin-based bind–elute processes are to acquire representative continuous poses several challenges. chromatography. Many experts starting materials and addressing ramp- Various industry groups have adopted anticipate that eventually all polishing up and shut-down phases as well as different approaches — including a flowthrough steps could be membrane- transient operation ranges of steady- tubular plug-flow reactor, a continuous adsorber–based. Current concerns about state unit operations. stirred-tank reactor, and column-based larger volumes could be addressed with Satzer: The first parameter to assess reactor — and successfully demonstrated new process schemes. Given that when moving to continuous operation is the implementation. chromatography steps comprise most of a new definition for exposure time. The challenge currently is a lack of the process bottlenecks and complexity Typically in batch processes this is standardized and validated viral in DSP, alternative approaches such as defined as a certain time period, but for inactivation strategy at scale or in a precipitation or flocculation that continuous operations there is a proven scale-down model. With increase product purity without residence-time distribution (meaning continuous processing, the significant compromising product quality/activity that some molecules stay longer, some problem is dividing an integrated before chromatographic steps will help not as long in the low-pH environment). operation into discrete unit operations streamline the process. It has to be ensured that even molecules for virus clearance testing. Virus Faude: Independent and continuous with shorter residence times will have filtration is a robust orthogonal method processing flowthrough operations are adequate viral inactivation. This issue in downstream applications that is used wherever possible to shorten can be prevented by implementing amenable for use in a continuous process times and achieve more larger safety margins or by using hybrid process. Issues that need addressing economic separations. Even for processes in which viral inactivation here are process variation/ chromatographies that traditionally remains a batch operation. perturbations and physicochemical were applied in bind–elute mode for Another approach is to redefine the feedstock changes. However, in both MAb purification (e.g., cation exchange), viral inactivation step itself. cases, validation-scale challenges are suppliers are beginning to provide new Historically, this is set (for antibodies) real and need to be understood before resins designed especially for to be around one hour. New research implementing a full-scale continuous flowthrough applications. shows that a few minutes are sufficient process. Alternative technologies are under for total viral inactivation. So in consideration, but platform processes comparison with the current approach, Residence Time are still in development. Technologies there already is a very large safety Can you describe the distribution of that can be used more generically will margin. I think the greater challenges residence time across a downstream attract greater interest. An example is in viral clearance has to do with process and how it affects products and the the recent trend toward considering filtration, because filters have to be bottom line? Is this a problem for flocculation technologies to optimize exchanged, and exchanges have to be continuous operations, or is there a way to harvest-filter capacity and precipitation validated. In general, technologies for address it? for optimizing impurity reduction. implementation of continuous viral Satzer: Residence-time distribution safety have been shown at laboratory through complete processes is poorly Virus Safety scale, so I think implementation is understood. Research has been limited Low-pH and detergent treatments are rather straightforward for this step. because batch processing does not have difficult to automate, which is necessary for Holzer: New technologies such the or need a description of residence-time continuous processing. Viral safety is not a BioSC pilot system (Novasep) allow for distributions. This question is unique to single unit operation, but rather is perfect control of unit operations such continuous processing and therefore expressed as the cumulative effects of as low pH- or detergent-based viral relatively new in the biopharmaceutical many operations. Does continuous inactivation, definition of different unit world. Models are available for some processing bring new challenges to operations (such as staggered-batch, unit operations and for some molecules, achieving such results? And which viral parallel-batch, or continuous but to fully understand all residence- safety operations will be most difficult to multicolumn chromatography), time distributions of all unit operations adapt to a continuous approach? integration of several different unit and impurities and products, we still Faude: Low-pH treatment especially operations to achieve cDSP, and the need a lot of research. To assess the needs equipment and technology for possibility for different process question for product quality, we have to appropriate process control. Detergent scheduling of simultaneously run unit define what components to track, treatment can be implemented easily operations. Control and integration of because not all have the same residence- from a technical point of view, but viral filtration into cDSP needs more time distribution. For instance, detergent removal, potential side development. aggregates tend to elute at the front and reactions with the detergents, and the Monge: Virus inactivation (VI) is back of a protein-A elution peak, potential presence of trace impurities considered one of the two orthogonal creating different residence-time increase development and analytical steps required for virus safety in a distributions for the product and for the costs. Additional challenges in virus biotherapeutic production process. impurities. Both have to be known clearance studies for continuous Making the process automated and before you can make any decision on

46 BioProcess International 17(6) June 2019

17-6-Satzer-RT.indd 46 8/28/2019 10:29:22 AM product quality at the end of your biopharmaceutical companies. In this process. We experience quite context, it would be better first to Only the generation of true “digital DRASTIC develop a batch-based platform process twins” with process models for all in and manufacture product for clinical, involved molecular species (such as CHANGES process validation, preapproval/ product, impurities) can solve this impurity patterns and registration, and commercial scale. That completely. But discussion has only just therefore downstream would enable companies to enter the started on what parameters have to be development when market and cater to its needs quickly tracked and what might be omitted from switching from batch- without having to worry about those models. uncertainties that continuous produced material to Monge: Processing time ranges need processing could bring from quality, to be characterized for unit operations perfusion material. manufacturing, automation, and and integrated downstream processes to regulatory perspectives. A parallel work ensure product stability and control of For antibodies, we know that protein stream needs to be in place where the impurities. Based on the criticality of A will perform, regardless of what same product is developed using a product residence time at a specific unit material you use, and minimal continuous process along with operation, additional studies might be adjustments are necessary when the evaluation of the transition from batch necessary to define control strategies feedstock changes. But for products that to continuous manufacturing and (e.g., tight control of pH and time for do not have a high-performance affinity identifying a bridging strategy. Once low-pH virus inactivation). The capture option, providing material that that is completed, then a postapproval characterization of residence-time adequately resembles that produced in change to a continuous process can be distribution becomes even more pilot/large-scale DSP as early as possible made by filing a prior approval important in cDSP because it serves also is crucial for process development. supplement (PAS) as outlined by the US as information for material traceability Holzer: Typically, process design of Food and Drug Administration (FDA) in and helps during impact analyses in countercurrent chromatography or its latest draft guidance covering quality case of investigations. membrane steps, plug–flow reactions, considerations for continuous in-line dilution (ILD), and so on need manufacturing. c Timing/Transition additional developments and dimensioning studies. Therefore, this It’s a general belief that the transition from S. Anne Montgomery is cofounder and editor in batch to continuous processing should be development stage would be perfect for chief of BioProcess International. Peter Satzer, made as early as possible in product implementation. The results of these PhD, is a senior scientist with the Austrian Center of development. If you consider the studies allow also for specifying Industrial Biotechnology (ACIB), Muthgasse 18, challenges that come with transitioning too processing equipment requirements. A-1190 Vienna AUT; +43 1 47654 79114; petersatzer@ acib.at; and lecturer at the University of Natural However, changes that mainly concern early and too late, where does the “sweet Resources and Life Sciences, Vienna, Muthgasse 18, spot” typically come between them? process scheduling and not development 1190 Vienna AUT, email [email protected]. Satzer: At the moment we face some of the unit operation to achieve cDSP issues in process development. In my could be implemented later in product/ opinion, the switch starts with the cell process development. In some cases, To share this in PDF or professionally printed line. Current cell lines are adapted for facility-fit challenges might be format, contact Jill Kaletha: jkaletha@ mossbergco. com, 1-574-347-4211. fed-batch high production, so they addressed with implementation of might not be at peak performance for in-line buffer dilution or continuous use in perfusion cultures. We experience multicolumn chromatography for one quite drastic changes in impurity single unit operation even for patterns and therefore downstream commercialized products. development when switching from Faude: The development of batch-produced material to perfusion continuous processing seems to be more material. time consuming unless you have a I think the earlier that transition is robust platform for some types of made, the better, and one limiting factor molecules. But that could conflict with (availability of small-scale equipment the importance of time to [reach] for running perfusion) has been lifted toxicology testing and time-to-market recently with combining the ambr considerations in early phases. The system (from Sartorius Stedim Biotech) transition to continuous processing with automated screening of might be during late-stage development chromatography performance using with a stronger focus on robustness and Robocolumns technology. Additionally, economical aspects. this might depend a great deal on the Monge: Speed to clinic and market product produced. are significant drivers for

48 BioProcess International 17(6) June 2019

17-6-Satzer-RT.indd 48 8/28/2019 10:29:22 AM