Truly continuous low pH viral inactivation for biopharmaceutical process integration.

Continuous virus inactivation (VI) has received little attention in the efforts to realize fully continuous biomanufacturing in the future. Implementation of continuous VI must assure a specific minimum incubation time, typically 60 min. To guarantee the minimum incubation time, we implemented a packed bed continuous viral inactivation reactor (CVIR) with narrow residence time distribution (RTD) for low pH incubation. We show that the RTD does not broaden significantly over a wide range of linear flow velocities - which highlights the flexibility and robustness of the design. Prolonged exposure to acidic pH has no impact on bed stability, assuring constant RTD throughout long term operation. The suitability of the packed bed CVIR for low pH inactivation is shown with two industry-standard model viruses, i.e. xenotropic murine leukemia virus and pseudorabies virus. Controls at neutral pH show no system-induced VI. At low pH, significant VI is observed, even after only 15 min. Based on the low pH inactivation kinetics, the continuous process is equivalent to traditional batch operation. This work establishes a concept for continuous low pH inactivation and, together with previous reports, highlights the versatility of the packed bed reactor for continuous VI, regardless of the inactivation method. This article is protected by copyright. All rights reserved.

30-60 min and neutralized with high concentration buffer. In mAB processing, low pH VI is often integrated with protein A chromatography, since its elution is performed at a pH-value close to those targeted for VI (Mattila et al., 2016). To avoid regulatory concerns about the possibility of hanging droplets or fluid inhomogeneities, a "two-vessel" strategy might be employed. In short, the process intermediate is acidified in a first vessel, transferred to a second vessel and only then the incubation time starts (Shukla & Aranha, 2015). Continuous low pH VI must assure that critical process parameters (CPPs) stay within prevalidated operation limits throughout the whole process. Typically these include, at least, pH-value, temperature, and minimum incubation time (Brorson et al., 2003;Mattila et al., 2016).
Temperature and pH-value can be easily controlled through thermostats and in-line mixers, respectively. Achieving the required minimum incubation is a more challenging feat, as a discrete incubation in batch translates into a residence time distribution (RTD). Recognizing the importance of RTD, the FDA draft guidance recommends RTD characterization for continuous processing (FDA, 2019). In strictly mathematical terms, the minimum incubation time for 100% of the fluid elements cannot be guaranteed given the statistical nature of the RTD determinations (also designated E-curve). Alternatively, a minimum residence time (MRT) approach has been suggested, in which 99% or 99.5% of the fluid elements are incubated for at least the target incubation time, for example, 60 min (David et al., 2019;Klutz, Lobedann, Bramsiepe, & Schembecker, 2016;Martins et al., 2019). To satisfy the MRT approach, the time at which the cumulative RTD (also designated F-curve) reaches, for example, 0.5% (t[F 0.5% ]) must equal 60 min-a typical minimum incubation time for low pH VI in batch mode.
For the specific case of continuous VI, the chance of any fluid element leaving the reactor before the target incubation time should be minimized, which highlights the requirement for a narrow RTD (Jungbauer, 2018). To achieve such a goal, three reactors have been patented and published. The coiled flow inverter (CFI) reactor (Klutz, Kurt, Lobedann, & Kockmann, 2015;Maiser, Schwan, Holtman, & Lobedann, 2016) and the jig-in-a-box (JIB) reactor (Orozco et al., 2017;Coffman, Goby, Godfrey, Orozco, & Vogel, 2015) rely on Dean vortices to narrow the RTD in coiled open tube. Dean vortices-a secondary flow pattern that provides axial mixing-are generated by a fine balance of centripetal forces and centrifugal forces over a narrow Reynolds number range (Parker et al., 2018). An alternative approach based on the packed bed was suggested . The bed of nonporous particles breaks the flow velocity profile thus resulting in a narrow RTD. Furthermore, the packed bed reactor outperforms the CFI and the JIB reactors of comparable scales in terms of RTD (Senčar, Hammerschmidt, Martins, & Jungbauer, 2020). Other systems have been suggested for continuous VI, however, some are not truly continuous in nature but rather cyclic (Gjoka et al., 2017;PALL, 2018) or do not consider the RTD when designing the incubation time (Arnold, Lee, Rucker-Pezzini, & Lee, 2019;Vaidya et al., 2018;Xenopoulos, 2013Xenopoulos, , 2015. The metric chosen for RTD narrowness measurement is also of great importance. Given the fact that peak fronting (i.e., early exit) is the greatest concern in a continuous VI process, the initial peak steepness is preferred for reactor assessment over other metrics that rely on the fitting of the whole F-curve, such as the Bodenstein number (Senčar et al., 2020). The initial peak steepness is the ratio of the time at which the cumulative RTD reaches 50% and a defined threshold percentage, for instance, 0.5% (t[F 50% ]/t[F 0.5% ]) and has become the preferred characterization method/metric for a reactor's performance (Klutz et al., 2016;Martins et al., 2019;Orozco et al., 2017). The choice of the threshold percentage must be small enough to describe peak fronting and early-existing fluid elements, but must be also large enough to be reproducibly monitored by commonly used detectors available for column performance testing and tracers generally regarded as safe (Amarikwa, Orozco, Brown, & Coffman, 2018). Overincubation should also be considered, especially for continuous VI designed according to the MRT approach where, by definition, most of the fluid elements are incubated longer for longer than the equivalent batch time. It is thus clear that the RTD analysis should also consider peak tailing (e.g., [F 99.5% ] has been used before [Klutz et al., 2015]). Overincubation might lead to product loss or damage, for instance, mAB aggregation at low pH (Joshi, Shivach, Kumar, Yadav, & Rathore, 2014). However, the published data for continuous VI systems do not suggest strong tailing in the RTD (Klutz et al., 2015;Orozco et al., 2017;Senčar et al., 2020) and such evaluation is the process-and product-specific . One common concern in continuous processing is performance deterioration, namely changes in CPPs, with consequences for product quality and safety.

| 1407
Besides assuring the CPP and robustness throughout the continuous operation, it is crucial to demonstrate low pH viral inactivation with relevant model viruses. Regulatory guidelines recommend the measurement of the VI kinetics described as "a biphasic curve in which a rapid initial phase is followed by a slower phase" (Committee for Proprietary Medicinal Products, 1996; International Conference on Harmonization, 1999). The same guidelines suggest different viruses that can be used based on cell-line susceptibility, historic record, and relevance. In 2018, a publication assessed viral clearance using continuous low pH VI using a straight tube reactor (Gillespie et al., 2018) but unfortunately, no quantitative measurement of the RTD (ideally, initial peak steepness) was provided. More recently, a report detailed the continuous low pH VI using the CFI reactor (David et al., 2019).
Besides the continuous inactivation, the report also covers the integration with the preceding unit operation (protein A continuous chromatography), which was achieved with a homogenization loop.
Aside from low pH inactivation, continuous solvent/detergent (S/D) treatment using a packed bed continuous viral inactivation reactor (CVIR) has been shown by our group .
In the present work, we extend the concept of the narrow RTD, packed bed CVIR and describe its application for low pH treatment. The CVIR is characterized and the impact of MRT based on the RTD is discussed. Analysis of bead and bed stability under acidic conditions is provided. Two industry-relevant model viruses are used (a) to control for any equipment-induced viral inactivation and (b) to assess the effectiveness of the CVIR for continuous low pH viral inactivation and show process performance equivalent to batch mode.

| Reagents
All chemicals were of analytical grade and were purchased from Merck Millipore (Germany), unless otherwise stated.

| Test item, test item buffer, and acid stock
The test item represents a generic process intermediate from a biopharmaceutical process. The test item consisted of 130 mM glycine (pH 7.0 ± 0.1), 8.84 ± 0.88 g/L human serum albumin. The test item buffer consisted of 130 mM glycine (pH 7.0 ± 0.1). A solution of 2 M glycine (pH 2.7) was used as acid stock to achieve the target pH-value (pH 3.7 ± 0.1).

| CVIR characterization and controls
The continuous VI system is depicted in Figure 1a

| CVIR characterization
Frontal analysis with a noninteracting tracer (2% [vol/vol] acetone in water) was performed to characterize the CVIR. The cumulative RTD (F-curve) was registered and the initial peak steepness (t[F 0.5 ]/t[F 0.005 ], specific for continuous VI) as well as traditional chromatography metrics (height equivalent to a theoretical plate [HETP], in µm, and asymmetry at 10% peak height) were calculated.
The HETP was calculated based on the moments method and was corrected for the system contribution in bypass without column. The linear velocity investigated ranged from 5 to 300 cm/hr (or from 0.167 to 10.1 ml/min, respectively). The asymmetry was calculated based on the derivative of the acetone front.
In this study, we designed and operated the reactor according to the MRT approach to guarantee that 99.5% of the fluid elements are incubated for the respective target time. More precisely, the definition of the reactor volume (V R ) relevant for continuous VI equals the volume at which the cumulative RTD .85 ml at 19.6 cm/hr (or 0.657 ml/min, the highest linear velocity/flow rate employed for VI studies and hence the worst case for fronting). Note that the (cumulative) RTD is typically shown as a function of time but can be shown as a function of volume by conversion with the volumetric flow rate. Throughout this study, the incubation times (t) in a continuous mode referred are calculated based on the volumetric flow rate, Q total , and the V R according to t = V R /Q total -this reflects the ever-present objective of assuring at least the respective incubation time. However, using the MRT approach also means that the majority of the fluid elements are incubated longer than the design criterion (t[F 0.5% ]).
For instance, in the case of the 15-min incubation experiments is 17 min-11.3% longer than the design criterion ( Figure 1b). This slight overincubation is a feature common to all continuous VI processes designed based on the MRT approach.

| CVIR stability under acidic pH
Bead swelling/shrinkage PMMA particles in different conditions (dry, in aqueous suspension at neutral pH and in aqueous suspension at pH 3.0) were analyzed in a phase-contrast microscope (Olympus CKX53SF with ×4 objective). The images were recorded using an Olympus SC50 digital image acquisition module (Olympus Europa, Germany). The beads were deposited on a six-well plate and analyzed on-plate under the different conditions to avoid sampling artifacts. For pH 3.0 incubation, a 50 mM sodium citrate, 150 mM NaCl (pH 3.0) buffer was used. The choice of pH 3.0 was selected as a worst-case scenario. Image processing (component analysis) was performed in Mathematica 11 (Wolfram) to identify circle-like shapes and their respective diameter was calculated. Each image/analysis was visually validated and images with more than one misidentified particle were excluded from statistical analysis.

Packed bed stability
A 130-mm HiScale 16 column was packed with 300-µm diameter PMMA beads and frontal analysis (with 1 M NaCl) was performed to determine initial peak steepness. Then, the CVIR was exposed for up to 10 days to 50 mM sodium citrate, 150 mM NaCl (pH 3.0) buffer. For this experiment, NaCl was chosen to avoid changes in concentration due to evaporation. The initial peak steepness was measured at different points during the exposure period. The

| RESULTS AND DISCUSSION
The CVIR was first characterized with respect to its RTD-this is the first step to implement a continuous VI process. The packed bed CVIR was additionally assessed regarding its stability towards prolonged low pH exposure, both in terms of particle size and packed bed RTD.  (Senčar et al., 2020).
Increasing the linear velocity to up to 300 cm/hr (or 10.1 ml) has a limited impact on initial peak steepness (see Figure 2a), which high-

| Particle stability at low pH
Considering the final intended application, the stability of the PMMA particles as well as of the packed bed itself was evaluated at pH 3.0 ( Figure 3). To assess PMMA bead swelling or shrinkage, multiple microscope fields were analyzed by an image processing script.
The result of a single analysis is shown in Figure 3a, where the software-identified particles are shown to match the actual particles.
(a) (b) (c) F I G U R E 2 (a) Column characterization by the t[F 0.5 ]/t[F 0.005 ] metric, (b) height equivalent to a theoretical plate (HETP) in µm, and (c) asymmetry at 10% peak height F I G U R E 3 Poly(methyl methacrylate) (PMMA) stability at acidic pH. Exemplary results of particle identification and determination by image processing. (a) The dotted circles denote the particle identified by the image processing tool. The scale bar is 500 µm. PMMA particle diameter probability density function (PDF) upon exposure to different conditions. (b-e) The gray dashed line represents the reference data (dry beads) and the solid black line represents the respective condition. The quartiles of the diameter distribution and the number of successfully identified particles are shown in the top, left inset. (f) PMMA packed bed stability upon exposure to pH 3.0. The initial peak steepness was measured five times per contact time point and the average ± standard deviation is shown To maximize the number of observations per condition, a compromise of one misidentified particle per field was accepte;, images with more than one misidentified particle were excluded from the analysis. Incubation of the particles in water for up to 45 hr (Figure 3b,c) did not change the particle size distribution when compared with the dry state. Similar results were obtained when the PMMA beads were incubated in citrate pH 3.0 for up to 7 days (Figure 3d,e). Also, only negligible changes in the particle diameter quartiles (d 25 , d 50 , and d 75 ) were found across different conditions. The initial peak steepness did not change considerably upon exposure to low pH (Figure 3f).

| Achieving the target pH
While in batch mode, the pH can be easily adjusted by the addition of concentrated acid and simultaneous pH measurement, in the continuous mode, this approach is not feasible. The simplest way to overcome this limitation is to preliminarily titrate the spiked test item with the acid stock to determine the volume of acid stock needed to reach the target pH ( Figure 4). An acid volumetric fraction of 2.38% is needed to acidify the test item to pH 3.7 ± 0.1, independently of the virus stock used. Titration of the spiked test item is dependent on the test matrix and acid stock used, and therefore it must be performed on a case-by-case basis (Gillespie et al., 2018) when the pH feedback control loop is not available (typically the case for small-scale process development and validation).
The pH-value was confirmed both at the reactor inlet and at the reactor outlet throughout the continuous operation ( Figure 5). The pH-value is a CPP in any low pH VI process, hence tight monitoring and control are required. As expected, the in-line mixer was effective in homogenizing the spiked test item and the acid stock to reach pH 3.7 ± 0.1. After the initial ramp-up phase, which lasts until 2 V R , the pH-value at the reactor outlet was constant and within the targeted range (pH 3.7 ± 0.1). The stable pH from 2 V R onwards is an indication that the system reached steady-state operation. Based on the inlet and outlet pH-values, it is fair to assume that the pH inside the column is within the target pH range. The use of inert, nonfunctionalized PMMA beads also supports this assumption, as there is no ion exchange between the fluid phase and the stationary phase inside the packed bed CVIR.  at 1 V R , the expected titer is 0.5% of the initial titer. However, the titers registered are 1.0-1.4 log 10 (TCID 50 /ml) above the expected titer. In practice, an infinitely small sample is not the case, but rather the sample covers a volume before and after 1 V R (or V[F 0.5% ]), so a titer above that based on the RTD is expected.

| CVIR-system controls
Additionally, a contributing factor is the sampling process, in which a sample is drawn at the CVIR outlet (~0.5 ml) and only part of if (0.2 ml) is diluted for neutralization and titrated, thus the excess of sampled material can contribute to the higher titer.
It is noteworthy, that the influence of the sampling process is only relevant for the ramp-up phase, as in steady-state the fluid elements before and after have the same composition and incubation time, leading to the same VI performance. That is the case of the samples at 2 V R and onwards, which show a constant titer, suggesting steady-state operation (as supported also by the RTD characterization, Figure 1b, which indicates that steady state is achieved after 1.3 V R , i.e., F 99.5% ). The lack of systeminduced VI is especially important for, and simplifies, process validation as any possible system-induced VI would have to be controlled for and the underlying phenomenon would have to be clarified (International Conference on Harmonization, 1999). The lack of system contribution means that the entirety of the VI can be attributed to the respective mode of action, in this case, exposure to low pH.

| Continuous VI and comparison against batch
The VI runs were performed at different flow rates to achieve 15, 30, and 60 min incubation (Figure 7a,b). For the model virus X-MuLV, the low pH viral inactivation was fast, resulting in no detectable infectivity even for the shortest incubation time tested (15 min). Full inactivation of X-MuLV was also registered after 14.5 min exposure to low pH using the CFI reactor (David et al., 2019). In the case of the PRV, extensive inactivation was also found in most of the samples drawn throughout the operation. Because the viral titer was already below (or at) the TCID 50 limit of detection, longer incubation times These results indicate comparability of the packed bed CVIR to the traditional batch operation-a likely regulatory concern.
4 | CONCLUSION A truly continuous low pH viral inactivation process was implemented using a narrow RTD-packed bed reactor. The same CVIR was characterized at linear velocities ranging from 4.90 to 300 cm/hr, corresponding to incubation times from 1 to 60 min.
It is important to highlight the packed bed system's flexibility with respect to incubation time, since increasing the linear velocity had limited negative impact on the initial peak steepness and RTD,

| OUTLOOK
The integration of continuous VI into a continuous process is a challenging task. Most protein A periodic counter-current chromatography (PCC) reports describe the use of three or four columns (Pollock et al., 2013;Warikoo et al., 2012;Zydney, 2016), which leads to a periodic and discontinuous outflow from the unit operation. To cope with a discontinuous mass flow, a surge tank might be necessary The mixer and packed bed (or any other narrow RTD) reactor will not dampen such fluctuations significantly. If the following unit operation is not affected by changes in product concentration and other potential CPPs, a continuous VI might be realized without surge tanks, providing that undisrupted mass flow is assured. If the following unit operation requires a constant-composition stream, then one surge tank is required either before or after the continuous VI. The size of the surge tank and concomitant RTD broadening should be the topic of careful consideration, as the large vessels would dampen fluctuations to a greater extent but would also propagate an out-of-specification case to a larger volume with its potential loss. Studies on the whole process's RTD are necessary and will be of invaluable help in such decisions.
The backpressure generated by one individual unit operation is also a topic of concern when integrating multiple unit operations. For instance, in the case of JIB, a pressure drop below 5 psi was cited as a design goal (Orozco et al., 2017). Contrarily to traditional chromatography, where pressures exceeding 5 psi are common, the packed bed CVIR is characterized by low back pressure due to (a) the larger particles used (200-400 µm in diameter) and (b) the lower linear velocities employed for continuous VI (e.g., 4.90-19.6 cm/hr). As an illustrative example, the Carman-Kozeny equation predicts a back pressure of 0.0586 kPa (or 0.00850 psi) at 19.6 cm/hr, which makes a reliable measurement impossible with the standard pressure monitors in typical chromatography workstations (Senčar et al., 2020).
Validation of a continuous VI is also a topic of concern. Besides inactivation studies with live viruses (inactivation kinetics), demonstration of stable process parameters might be also required. While parameters like pH-value or temperature can be easily measured, measuring the RTD throughout the operation is not easily performed.
In a continuous VI processes designed according to the MRT approach, the RTD (or at least V[F 0.5% ]) could be viewed as a CPP. How this question is addressed depends on the system used to provide continuous incubation. Recently, Brown et al. (2019) published a datadriven approach to characterize the impact of viscosity on the Dean number/RTD narrowness and provide an empirical relationship for the JIB scale up. Our group has characterized extensively the packed bed reactor and how a feature like a particle size, linear velocity, or packed bed geometry contribute to the initial peak steepness (t[F 0.5 ]/t[F 0.005 ]) (Senčar et al., 2020). Despite the data produced, it is likely that RTD measurement of the actual reactor employed in manufacturing would still be required as a performance test both before and after use (similarly to an integrity test for a viral filter). Such an approach implies that a GMP-compliant tracer must be used, thus limiting how low the threshold in the MRT approach can be.

ACKNOWLEDGMENTS
This study has been supported by the Federal Ministry for