- Email: [email protected]
A narrow residence time incubation reactor for continuous virus inactivation based on packed beds
New BIOTECHNOLOGY 55 (2020) 98–107
Contents lists available at ScienceDirect
New BIOTECHNOLOGY journal homepage: www.elsevier.com/locate/nbt
Full length Article
A narrow residence time incubation reactor for continuous virus inactivation based on packed beds Jure Senčara, Nikolaus Hammerschmidta, Duarte L. Martinsa, Alois Jungbauera,b,* a b
Austria Centre for Industrial Biotechnology, Vienna, Austria Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria
ARTICLE INFO
ABSTRACT
Keywords: Residence time distribution Log reduction value Continuous operation Non-porous inert beads Downstream processing Plug flow
A narrow residence time distribution (RTD) is highly desirable for continuous processes where a strict incubation time must be ensured, such as continuous virus inactivation. A narrow RTD also results in faster startup and shut down phases and limits the broadening of potential disturbances in continuous processes. A packed bed reactor with non-porous inert beads was developed to achieve narrow RTDs. The performance was defined as the ratio between the onset of the cumulative RTD and the median residence time (tx%/t50%). Laboratory-scale packed columns were used to study the influence of the column parameters on the RTD. A larger column with a void volume of 0.65 L and a length of 89 cm, packed with beads in a size range of 125 to 250 μm, achieved t0.5%/ t50% > 0.93 across flow rates from 0.1 to 9.8 mL/min. The RTD was significantly narrower than the RTDs of other reactor designs, such as the Coiled Flow Inverter and Jig in a Box. The pressure drop remained under 3 kPa for all tested flow rates. Fluorescent nanoparticles (30 and 200 nm) were used to mimic viruses. These two sizes showed less than 2% difference in terms of t1%/t50% and t0.01%/t50% scores. These results indicated that viruses travelled through the column at rates independent of size. This proposal of packed beds as incubation chambers for continuous virus inactivation is simple, scalable, and can be realized as single-use devices. Due to the low pressure drop, the system can be easily integrated into a fully continuous process.
Introduction In the development of continuous integrated biopharmaceutical processes, one of the key challenges is continuous virus inactivation, because a defined, strict incubation time must be ensured [1]. This can be achieved with either a semi-continuous approach, which alternates between two batch reactors, or with a fully continuous approach, where an ideal plug flow reactor is desirable [2]. Continuous biomanufacturing is interesting, due to the potential economic benefits, such as the reduced unit operation size, the reduced footprint, and the possibility of working with disposables, even at full scale [3–5]. To render a process continuous, all unit operations need to be operated in a continuous or semi-continuous manner. A narrow residence time is required to ensure a given reaction time, but also to
limit the time for starting up and shutting down. The latter point is of particular interest when the process is interrupted. Regulatory agencies demand two orthogonal virus inactivation/removal steps for cell-culture derived biopharmaceuticals. The first step is the physical removal of viruses with nanofiltration [1]. The second requires diverse methods that depend on the process and product. The commonest ways to inactivate viruses in processing intermediates include low pH inactivation [6], solvent-detergent treatment [7], and ultraviolet C light inactivation [8]. All of these approaches work by exposing the process fluid to conditions detrimental to viruses, and incubating them in the condition for a given time. In the context of continuous operation, the batch incubation time (value) becomes a time distribution. Operated in batch, low pH inactivation and solvent-detergent
Abbreviations: λ, wavelength; t, residence time; tx%, residence time at which the normalized signal reaches x %; tbatch, batch incubation time; tmean, mean residence time; tLOD, residence time at which the signal reaches the limit of detection; FLOD, cumulative RTD at which the signal reaches the limit of detection; LOD, limit of detection; LRV, log reduction value for virus inactivation; LRV batch, required log reduction value for virus inactivation; RV, reduction value for virus inactivation; RV continuous, minimum guaranteed reduction value for a continuous virus inactivation; RV LOD, reduction value for virus inactivation for tLOD; RV preLOD, reduction value for virus inactivation for before tLOD is reached; RTD, residence time distribution; CFI, Coiled Flow Invertor; CV, column volume; JIB, Jig in a Box; PMMA, polymethylmethacrylate ⁎ Corresponding author at: Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria. E-mail address: [email protected] (A. Jungbauer). https://doi.org/10.1016/j.nbt.2019.10.006 Received 15 October 2018; Received in revised form 8 October 2019; Accepted 9 October 2019 Available online 17 October 2019 1871-6784/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
treatment are performed with exposure times of at least 30 min, but typically 60 min [9,10]. Moving towards continuous operation, a narrow residence time distribution (RTD) is of paramount importance to ensure sufficient incubation time and to prevent excess exposure of the product to harsh virus inactivation conditions. Multiple concepts for continuous or semi-continuous operation have been proposed in recent years. A semi-continuous approach towards continuous virus inactivation that employs two alternating tanks, which mimics a batch operation is currently on the market (Cadence™ Virus Inactivation system, Pall Corporation). A fully continuous virus inactivation system, based on UV-C radiation, is also currently on the market (UVivatec™, Sartorius Stedim Biotech). Recently, a fully continuous virus inactivation system that employs a size exclusion (SEC) column as the incubation reactor was reported [11]. However, those authors did not discuss the RTD or the characteristics of the reactor. Other fully continuous virus-inactivation reactor designs, with long residence times, have been reported recently in the literature, such as the Coiled Flow Inverter (CFI) [12] and the refined Jig in a Box (JIB) concepts [13,14]. Those designs were based on the induction of secondary radial flow patterns, which narrowed the RTD, compared to straight tubes. The efficiency of those reactors varies with the fluid velocity and tube diameter. The tubes employed in those types of reactors are typically very long relative to their diameter, to achieve a narrow RTD. However, that design can lead to back-pressure limitations at larger scales. A possible continuous virus inactivation setup (Fig. 1) would start with a static or dynamic inline mixer for adding the inactivation chemicals to the process fluid, and an inline sensor, such as a pH probe or a Fourier transform infrared detector, to ensure conditions remain within defined limits. The mixture then enters the incubation chamber. At the outlet of the incubation reactor, there is another inline mixer for neutralizing virus-inactivation chemicals. A 0.22 μm dead-end filter is placed before and after the inactivation chamber to control the bioburden and to remove aggregates, which could form before or during the incubation [1]. In the present study, an incubation reactor for a fully continuous virus inactivation is developed, based on well-established packed beds of non-porous inert beads. Beads with diameters between 100 μm and 1 mm were considered to achieve a narrow RTD and to avoid high backpressure and size-exclusion effects (i.e., different flow-through times for larger and smaller viruses). There was no need for custom-designed parts, because well-established chromatographic equipment was already available. In addition, the influence of design parameters on the RTD, such as column dimensions, bead size distribution, operating flow rate and the diameter of tracer nanoparticles, was investigated.
Materials and methods Equipment and buffers Chromatographic columns HR 5, HR 10, HR 16, HS 16, and XK 50 (all GE Healthcare) were packed with non-porous beads. Glass beads (Sigmund Lindner GmbH) of different diameter distributions (0.1 mm – 0.2 mm; 0.2 mm – 0.3 mm; 0.3 mm – 0.4 mm; 0.25 mm – 0.5 mm; and 1.0 mm – 1.3 mm) were used to study the influence of the bead size distribution on column performance. We tested columns with diameters in the range of 5–16 mm and heights between 3.8 and 29.3 cm. Inert beads composed of several different materials were considered for the final continuous virus inactivation application (Table 1). Based on inertness, price, sphericity, density, and robustness, polymethylmethacrylate (PMMA) beads were selected as the most appropriate. A new column was packed with PMMA beads and added to the group of glass-bead columns (Supplementary Table 1). All RTD characterization experiments were performed on an Äkta Avant™ (GE Healthcare) chromatography system. The outlet concentration was measured by the UV cell of the system. Breakthrough experiments were performed by equilibrating the column with reverse osmosis-H2O (RO-H2O), and then switching to a 2% acetone solution. The breakthrough profiles were obtained by measuring the acetone concentration at the outlet of the column, at λ =280 nm. A solvent-detergent mixture/physiological buffer system was used to investigate potential differences in column performance compared to the RO-H2O/acetone system. The solvent-detergent mixture was prepared with final concentrations of 1% Triton X-100, 0.3% Polysorbate 80, and 0.3% Tri-n-butyl phosphate, in the physiological buffer. The content of solvent-detergent chemicals was measured at λ =300 nm. To simulate viruses of various sizes traveling through the column, florescent, virus-sized nanoparticles (Micromod Partikeltechnologie GmbH, Sicastar®-greenF, plain, excitation: 485 nm, emission: 510 nm) of two different diameters (30 nm and 200 nm) were used. A short pulse (< 0.5% of the column volume) of nanoparticles (25 mg/mL) in an aqueous suspension was injected at various operating flow rates. An UltiMate™ 3000 Fluorescence Detector (Thermo Scientific) was used to detect the fluorescent nanoparticle tracers at the outlet of the column. The peak obtained was integrated, and then evaluated as a breakthrough profile. Column packing Column packing was performed with a custom-built vibrating column holder. The column was filled with water. Then the column was set to vibrate, as beads were slowly poured in at the top of the column. Column vibration continued, as they immersed in the water. Then, the column was closed and tightened, before vibration was stopped. Gravimetric testing of the packed columns showed that porosity varied by < 2% for columns packed with the same batch of beads. During 8 months of column usage, we observed no shrinkage of the bed height or decrease in column performance, based on measurements taken in periodic breakthrough experiments, including a prolonged exposure to 1 M NaOH and various aqueous buffers. Breakthrough profile characterization Peak-fronting is a primary concern in continuous virus inactivation, because the earliest volume fractions have the shortest incubation times. The narrowness of the RTD was evaluated as the ratio between the time that the concentration of the breakthrough buffer reached a certain small percentage (tx%) and the median RTD (t50%) (Fig. 2a). The closer the tx% /t50% ratio is to 1, the closer the onset of a peak is to an ideal plug flow. The Bodenstein number was obtained by fitting a breakthrough curve to the F function, as described previously [12]. In our experience, the F function does not provide a good fit to the
Fig. 1. A diagram of a potential continuous virus inactivation setup. An inline mixer continuously mixes the virus-inactivation chemicals with the process fluid. Next, the mixture passes through a 0.22 μm dead-end filter and the inline sensor (e.g., pH probe or Fourier transform infrared detector), which ensures consistent mixture composition. The mixture then flows through the incubation reactor, and after incubation, it is routed through the second inline filter. The virus inactivation properties of the process fluid are then neutralized by mixing in neutralization chemicals via a second inline mixer. 99
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Table 1 Comparison of the different types of inert beads considered for this study. Bead material (manufacturer)
Particle size distribution
Density (kg/L)
Cost
Sphericity
Glass (Sigmund Lindner GmbH) PMMA (Kisker Biotech GmbH) Ceramic (Kuhmichel Abrasiv GmbH) Polystyrene (Cospheric LLC) Polyethylene (Cospheric LLC)
broad medium medium narrow narrow
2.7 1.18 3.8 1.04 0.97
low medium low high high
medium very high high very high very high
PMMA: polymethylmethacrylate.
experimental data. Furthermore, it is not sensitive to the peak onset (Fig. 2 b). Therefore, because the peak onset is crucial for claiming a sufficient log reduction value (LRV), it was decided to use the tx%/t50% ratio as a metric for the narrowness of the RTD throughout this work. A similar concept for evaluating the narrowness of the RTD was previously employed by others [13].
performed on the column packed with PMMA beads. The tested superficial linear velocities were 3 cm/h, 9 cm/h, 30 cm/h, and 120 cm/h. Again, higher flow rates were not tested, due to the high back-pressure generated by the analytical fluorescence detector.
Investigating column parameters
Claiming the log reduction value
To investigate the influence of column parameters on the RTD, twelve columns packed with glass beads of various bead sizes and one packed with PMMA beads were characterized. Acetone breakthrough experiments were perfomed at superficial linear velocities of 50, 100, 200, and 300 cm/h. To confirm the narrowness of the RTD after scaling-up, a larger XK50 column (diameter =5 cm, height =89.0 cm, bead size = 125–250 μm) was packed and the RTD tested at superficial linear velocities that ranged from 5 cm/h (1.6 mL/min) to 30 cm/h (9.8 mL/ min). Higher flow rates were not possible, because the system pressure, generated by the UV detector, the tubing, and the valves, approached the specified pressure limit for the low-pressure XK50 column frame (< 0.3 MPa). Experiments with virus-sized fluorescent tracer nanoparticles were
Two methods have been suggested recently for achieving batchequivalency with regard to virus removal [15]. The minimum residence time approach was to set the incubation time to ensure that 99.5% of the peak eluted no earlier than the required batch incubation time (e.g. 60 min). The second approach had a more theoretical nature; it suggested selecting the incubation time to ensure that the weighted average of the LRV across the RTD was the same as the required LRV in a batch operation. Note that, in this approach, a larger portion of process fluid was allowed to have a shorter residence time than the batch incubation time. However, that approach requires a detailed analysis of virus inactivation kinetics and the exact RTD of the continuous virus inactivation reactor. It was decided to extend the minimum residence time approach by focusing on the very beginning of the RTD – even before the lower limit
Results and discussion
Fig. 2. Measurements of the narrowness of the residence time distribution (RTD) for continuous virus inactivation. (a) The narrowness is defined as the ratio between the time at the onset, when a small percent (e.g. t5%, dashed line) of the cumulative RTD has flowed through, and the median RTD (t50%, dot-dashed line). (b) The t1%/ t50% values and the Bodenstein numbers (Bn) were calculated for sample experiment breakthrough profiles with various shapes. The Bn values were similar (408 and 452) between two breakthrough profiles with similar overall shapes, but different peak fronts. In contrast, the t1%/t50% values showed better discrimination between the profiles. Thus, the t1%/t50% ratio is more sensitive to peak fronting, which is a critical feature for ensuring sufficient virus inactivation. Abbreviations: CV: column volume.
100
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
of detection (LOD) was reached. For this, a theoretical worst-case scenario was defined, based on the requirements for the batch process. A logarithmic nature for the virus inactivation kinetics is assumed. In fact, though, inactivation of virus infectivity frequently follows a biphasic curve, where a rapid initial phase is followed by a slower phase [1]. Thus, compared to a purely logarithmic function, virus inactivation typically displays much faster kinetics at the very beginning of the incubation period and slower kinetics at the end [2,6,9]. Therefore, the assumption of a logarithmic nature added an extra margin of safety to the approach. A reduction value (RV) at any time (t) was defined as the ratio between the current virus concentration (c) and the initial virus concentration (c0). The assumed virus inactivation kinetics for the batch process, with its required LRV (LRVbatch) and its required incubation time (tbatch), was described with Eq 1.
RV (t ) =
c (t ) = 10 c0
LRVbatch t tbatch
(1)
The LOD for the breakthrough profile is reached at the time (tLOD) when the normalized concentration of the species breaking through reaches a certain value (FLOD). By further assuming a single peak in the RTD, a worst-case envelope for the breakthrough profile is obtained, with a known FLOD and tLOD. The envelope of the assumed cumulative RTD rises linearly from t = 0 to tLOD, and reaches a value of 1 at tLOD (Fig. 3a). The envelope could be optimized by incorporating the t50% parameter (Fig. 3b). However, in our experience, the benefits of that optimization did not justify the additional complexity, when dealing with a very narrow RTD. The average virus RV before reaching the tLOD (RVpreLOD) was described with Eq. 2.
Fig. 4. A diagram of the worst-case theoretical scenario for a continuous virus inactivation process. We assumed a logarithmic nature of virus inactivation and a single peak in the residence time distribution. The X-axis is the required log reduction value (LRV) for an equivalent batch operation. The Y-axis is the limit of detection (LOD) for the normalized breakthrough profile (cumulative residence time distribution). The values represent the ratio of the minimum required time when the LOD is reached (tLOD) to the equivalent batch incubation time (tbatch).
Fig. 3. A worst-case assumption for a normalized breakthrough profile. (a) The assumption was based on the time point (tLOD), at which the breakthrough profile (cumulative residence time distribution) crosses the lower limit of detection (LOD). (b) The worst-case assumption can be optimized by adding information about the median residence time. In both cases, the worst-case envelope (dashed line) will always lie above the experimental data (solid line); thus, it can be safely used to approximate the normalized breakthrough profile with an additional safety margin.
101
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
RVpreLOD =
1 tLOD
tLOD 0
10
LRVbatch t tbatch dt
=
1
10
finding indicated that the input parameters were not strongly correlated over the selected range, and thus, they were suitable for a linear prediction model. The verification of the data yielded a clear trend between the measured and predicted values (Fig. 5b).
LRVbatch tLOD tbatch
LRVbatch tLOD tbatch
ln(10)
(2)
For the volume fractions after tLOD, the same RVLOD as that observed at tLOD was assumed [Eq. 3].
RVLOD = RV (tLOD ) = 10
LRVbatch tLOD tbatch
Buffer system with solvent-detergent chemicals A subset of experiments described in the previous section was repeated with the physiological buffer and the physiological buffer combined with the solvent-detergent chemicals as tracer. The results matched those from the same experiments performed with water and 2% acetone (Fig. 6). This finding justified the use of water and 2% acetone for the model buffer system. The latter buffer system was used for all subsequent experiments to achieve higher sensitivity and avoid using toxic chemicals.
(3)
The equation for the combined virus RV (RVcontinuous) was obtained by combining the reduction observed before the LOD and the reduction observed after the LOD [Eq. 4]. RVcontinuous = FLOD RVpreLOD + (1 – FLOD) RVLOD
(4)
The RVcontinuous should match the required RV for the batch operation, RVbatch [Eq. 5]. RVcontinuous = RVbatch = RV(tbatch) = 10 –LRVbatch
A column packed with PMMA beads Another analysis was performed with PMMA beads, because they were the most promising material for the intended application. Their density of 1.18 mg/mL allows for wet packing and keeps the overall weight of the column relatively low. PMMA particles were coated with a gold layer and subjected to scanning electron microscopy. Images revealed that the PMMA beads had much less surface irregularity than glass beads. In addition, no cavities or pores were detected (data not shown). The size distribution of PMMA beads was measured with a Morphology™ G3 instrument (Panalytical GmbH), and the results were within the manufacturer’s specifications (more than 95% of beads were within the specified 200–300 nm size range). The stability of the column packed with PMMA beads was also tested by prolonged flushing with 0.5 M NaOH (pH 13.7), to simulate the column regeneration/sanitization cycles, and with low pH buffer (50 mM citrate, pH 3.0), to simulate a potential low-pH inactivation mode. These chemical treatments did not alter the breakthrough profiles. The asymmetry of the column ranged from 1.1 to 1.3, depending on the superficial linear velocity (Fig. 7a). A linear log-log dependency was observed between height equivalent to a theoretical plate (HETP) and the superficial linear velocity (Fig. 7 b), consistent with an empirical model described previously [18].
(5)
By combining Eq.s 1–5, the required worst-case tLOD was calculated, based on batch process requirements (tbatch, LRVbatch) and the limit of detection (FLOD) [Eq. 6].
10
LRVbatch
= FLOD
1 10 LRVbatch
t LRVbatch t LOD
batch
tLOD tbatch
ln(10)
+ (1 – FLOD )* 10
t LRVbatch LOD tbatch
(6) The solution for Eq. 6 is presented in Fig. 4, in the form of a tLOD/ tbatch ratio depending on the LRVbatch and FLOD. The ratio approached unity as the LOD approached zero. A LOD that approached zero, combined with an ideal plug flow would cause the required mean residence time to approach tbatch. For a LOD = 0.01% and a required LRV = 5, the tLOD⁄tbatch = 1.13. In the case of a batch equivalent incubation time of 60 min, the tLOD = 68 min for a continuous virus inactivation, i.e. 68 min was the time required for the signal of the breakthrough profile to cross the lower LOD. The additional 8 min was needed to compensate for the equipment LOD, given the required LRV. A LOD of 0.01% may not be realized with existing in-line technology for any give process. Thus, an alternative approach might be needed, such as high sensitivity offline characterization. In our experience, it was possible to achieve a LOD of 0.01% with a 5% acetone solution; moreover, a LOD as low as 0.0001% was achieved in the cumulative pulse response of fluorescence nanoparticle tracers. Several potentially process-compatible tracers (that is, generally recognized as safe) were previously used for an offline characterization of JIB [16]. With the calculation that assumed the worst-case scenario for the continuous process, including experimental virus inactivation data from an established batch process, it was possible to claim an LRV at least as high as that achieved in the batch process.
Virus-sized nanoparticles The column packed with PMMA beads was also used for experiments with virus-sized fluorescent nanoparticles to investigate whether particles in the size range of viruses might travel through packed beds differently depending on their size. Fluorescent particles were used in combination with a fluorescence detector, due to the higher sensitivity compared to UV detectors. The RTDs obtained at different flow rates were evaluated in terms of the median residence time (t50%), the t0.5%⁄t50% ratio, and the t0.01%⁄t50% ratio. The difference in the results between the tracer sizes (30 nm and 200 nm bead diameters) was < 2% across the entire range of tested flow rates (Fig. 8). This finding was expected, because the bead diameters in the packed bed were much larger than the nanoparticles. An example was considered where an LRV of 3 is targeted using the column packed with PMMA beads. If a tracer with a LOD of 0.5% is used for reactor characterization, the tLOD/tmean = t0.5%/t50%. Thus we obtained a tLOD⁄tmean = t0.5%⁄t50% ≥ 0.78, based on Fig. 8, and a tLOD⁄tbatch = 1.15, based on Fig. 4. Based on Eq. 7 and a 60 min batch incubation time (tbatch =60 min), the required residence time was 89 min. To claim a LRV of 5 while using a LOD of 0.5%, the tLOD⁄tbatch value would increase to > 7, which would make the reactor impractically large. Thus, in cases that require a LRV of 5, the required LOD would be 0.01%, which gives tLOD⁄tbatch = 1.13 and tLOD⁄tmean = t0.01%⁄t50% ≥0.68 for the column packed with PMMA beads. These parameters
Influence of column parameters on RTD Twelve columns packed with glass beads and one column packed with PMMA beads were characterized at different linear velocities. The ratio t1%/t50% was correlated to column properties with a partial least square (PLS) model that had four components. The bead size distribution did not have a major influence on column performance, as found previously by others [17]. Thus, this property was excluded from the model. In the range tested, the average bead diameter had the largest influence, followed by the column length, the ratio between the column cross-section and the column length, and the flow rate (Fig. 5). The individual factors influenced the RTD (t1%⁄t50%) in the expected direction. However, no theoretical model available in the literature could describe the results quantitatively. The first PLS component alone explained 94% of the variability in the measurements (Fig. 5a). This
102
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 5. A four-component partial least square (PLS) model was employed to describe the influence of column properties on the RTD, in terms of the t1% /t50%. (a) The first component of the model (x) is compared to the corresponding linear regression model (bars). len: column length; A: column cross-section; vs: superficial linear velocity; dp: mean bead diameter. (b) Verification of the regression with the first component of the PLS model yields a trend with R² = 0.936.
time in an actual application would be expected to be closer to the batch incubation time.
tmean = tbatch *
tLOD tmean t t * = tbatch * LOD * 50% tbatch tLOD tbatch t x %
(7)
Backpressure In all columns examined, the pressure dropped below the LOD of the pressure sensor on the Äkta Avant system (< 3 kPa). Low backpressure is a desirable feature, because when multiple unit operations are directly coupled in continuous processing, the individual pressure drops add up. Fig. 9 summarizes the backpressure calculations with the Carman-Kozeny equation, assuming a porosity of 0.4 and a bead sphericity of 0.9. The equation is applicable to the examined system, because we used rigid, non-compressible particles. Fig. 9a illustrates the expected pressure drops for 250-μm particles with various residence times and column lengths. In Fig. 9b, the pressure drop across packed beds was calculated for various bead diameters and column lengths. Scaling-up To investigate the packed column performance at a larger scale, an XK 50 column was packed (diameter =5 cm, height =89.0 cm, bead size = 125–250 μm). The XK 50 column achieved a t5%/t50% value of > 0.93 across the entire tested range of linear velocities (Fig. 10). This finding indicated that this column achieved substantially narrower RTDs than the RTDs of other, smaller columns. To claim a LRV of 3, the parameters for Eq. 7 would be: tLOD⁄tmean = t0.5%⁄t50% > 0.93 and tLOD⁄tbatch = 1.15. For a batch equivalent incubation time of 60 min, the required mean residence time for the packed bed column would be 74 min. This time was significantly shorter than that attained with the much smaller column (tmean =89 min). However, to claim a LRV beyond 5 would need a lower LOD. This further emphasized the need for high-sensitivity process-compatible tracers. The CFI and JIB concepts were also implemented at larger scales; however, the Supradex 200 column [11] could only be used at a
Fig. 6. A comparison of residence time distributions (RTD) among the solventdetergent (SD) buffer system and the acetone buffer system. The RTD was evaluated in terms of the t5%/t50% ratio for various columns at various linear velocities. Individual data points correspond to two experiments performed under the same conditions, with different buffer systems. The linear trend justifies the use of the acetone buffer system as a model for the SD buffer system. Buffer systems: acetone buffer system: RO-H2O with 2% acetone as tracer; SD buffer system: physiological buffer with SD chemicals (1% Triton X100, 0.3% Polysorbate 80, and 0.3% Tri-n-butyl phosphate) as tracer.
resulted in a required mean residence time of 100 min for a batch equivalent time of 60 min. Although this residence time was significantly greater than the batch equivalent incubation time, it represented a theoretical worst-case scenario, and thus, the residence 103
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 7. Properties of a column packed with PMMA beads (d = 16 mm, h = 13.2 cm, bead size = 200–400 μm). (a) Asymmetry across superficial linear velocities. (b) A log-log relationship between superficial linear velocity and HETP. Abbreviations: PMMA: polymethylmethacrylate; HETP: height equivalent to a theoretical plate.
very low flow rate (< 1 mL/min) and it is not suitable for scaling up, due to the high backpressure generated by the SEC matrix.
performance of the CFI [12] and the JIB [13,14] is shown in Fig. 11. In terms of a narrow RTD, the smaller set of packed bed column reactors performed similarly to the CFI with much larger void volumes. Indeed, the larger packed bed column, with a void volume of 0.65 L, outperformed all reported CFI designs (void volumes up to 471 mL) and JIB designs with up to three JIB boxes in series (total void volume of 1.5 L).
Comparison with other reported designs The comparison of the packed bed performance with the
Fig. 8. Comparison of the properties of fluorescent particles with 30 nm and 200 nm diameters measured across different superficial linear velocities. (a) The median residence time, (b) the onset of the peak, in terms of t0.5% /t50% and (c) the high-sensitivity onset of the peak, in terms of t0.01% /t50%. The tx% is the residence time at which the cumulative residence time distribution reaches value of x%.
104
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 9. Expected pressure drops for columns packed with non-porous inert beads, based on the Kozeny–Carman equation. Pressure drops calculated for (a) columns of different lengths, packed with 250-μm beads, and run with various residence times; (b) columns of different lengths, with a mean residence time of 80 min, loaded with various bead diameters. For the calculations, we used a porosity value of 0.4 and sphericity value of 0.9.
Conclusion
inactivation methods, such as solvent-detergent treatments. The packed bed column attributes were correlated with the width of the RTD in a way that enabled the linear prediction of column performance. For virus-sized fluorescent particles (30 and 200 nm) it was demonstrated that the narrowness of the RTD and the median residence time did not depend on particle size (variability < 2%). This finding suggested that the passage of nanometer-sized particles through the reactor was independent of their size. The packed bed column significantly outperformed other designs,
We developed an incubation reactor based on packed beds of nonporous, inert beads for continuous virus inactivation, which offered a very narrow residence time distribution. The approach described combines existing, well-established bead chromatographic equipment and theory with a simple design and a disposable setup. The proposed system can be used for low pH inactivation, which is typically applied in monoclonal antibody downstream processes, and other virus
Fig. 10. Up-scaled processing with the XK 50 column (d =5 cm, h =89.0 cm, bead size = 125–250 μm) across different superficial linear velocities. (a) The median residence time; (b) the narrowness of the residence time distributions, in terms of the t1%/t50% ratio and (c) the t0.5%/t50% ratio, and (d) the Bodenstein number. Abbreviations: CV: column volume. 105
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 11. The performance of the packed bed described in the present study compared to the performance of other designs at different scales (void volumes). The packed bed is compared to (a) the Coiled Flow Inverter design [12] and (b) the Jig in a Box design [13]. In both plots, the larger symbols translate to a narrower RTD, (a) in terms of the Bodenstein number and (b) in terms of the t0.5% /t50% ratio. All designs achieved a narrower RTD at larger scales. At similar scales, the packed bed outperforms the other two designs. The t0.5% /t50% value range in plot (b) is between 0.85 and 0.97, where the ideal plug flow would have a value of 1.
such as the JIB and CFI which had up to 2-fold larger void volumes. The results also suggested that the RTDs of all three reactor designs became narrower with increased reactor size. The design did not rely on secondary flow, such as Dean vortices; therefore, it could be readily scaled up. The scale-up could be performed by applying a wide flow regime, and thus, the process would remain well within the laminar flow region. Backpressure, an important and potentially limiting parameter in integrated processes, was particularly low for the proposed bead sizes (> 200 μm) across all the experiments.
[2] Gillespie C, Holstein M, Mullin L, Cotoni K, Tuccelli R, Caulmare J, et al. Continuous in-line virus inactivation for next generation bioprocessing. Biotechnol J 2019. https://doi.org/10.1002/biot.201700718. [3] Jungbauer A. Continuous downstream processing of biopharmaceuticals. Trends Biotechnol 2013. https://doi.org/10.1016/j.tibtech.2013.05.011. [4] Rathore AS, Agarwal H, Sharma AK, Pathak M, Muthukumar S. Continuous processing for production of biopharmaceuticals. Prep Biochem Biotechnol 2015. https://doi.org/10.1080/10826068.2014.985834. [5] Shukla AA, Gottschalk U. Single-use disposable technologies for biopharmaceutical manufacturing. Trends Biotechnol 2013. https://doi.org/10.1016/j.tibtech.2012. 10.004. [6] Mattila J, Clark M, Liu S, Pieracci J, Gervais TR, Wilson E, et al. Retrospective evaluation of Low-pH viral inactivation and viral filtration data from a multiple company collaboration. PDA J Pharm Sci Technol 2016. https://doi.org/10.5731/ pdajpst.2016.006478. [7] Horowitz B, Prince AM, Hamman J, Watklevicz C. Viral safety of solvent/detergenttreated blood products. Blood Coagul Fibrinolysis 1994. https://doi.org/10.1097/ 00001721-199412003-00006. [8] Bae JE, Jeong EK, Il Lee J, Lee JI, Kim IS, Kim JS. Evaluation of viral inactivation efficacy of a continuous flow ultraviolet-C reactor (UVivatec). Korean J Microbiol Biotechnol 2009. [9] Brorson K, Krejci S, Lee K, Hamilton E, Stein K, Xu Y. Bracketed generic inactivation of rodent retroviruses by low pH treatment for monoclonal antibodies and recombinant proteins. Biotechnol Bioeng 2003. https://doi.org/10.1002/bit.10574. [10] E3042-16. Standard practice for process step to inactivate rodent retrovirus with Triton x-100 treatment. West Conshohocken, PA: ASTM International; 2016. https://doi.org/10.1520/E3042-16. [11] Arnold L, Lee K, Rucker-Pezzini J, Lee JH. Implementation of fully integrated continuous antibody processing: effects on productivity and COGm. Biotechnol J 2019. https://doi.org/10.1002/biot.201800061. [12] Klutz S, Kurt SK, Lobedann M, Kockmann N. Narrow residence time distribution in tubular reactor concept for Reynolds number range of 10-100. Chem Eng Res Des 2015. https://doi.org/10.1016/j.cherd.2015.01.003. [13] Orozco R, Godfrey S, Coffman J, Amarikwa L, Parker S, Hernandez L, et al. Design, construction, and optimization of a novel, modular, and scalable incubation chamber for continuous viral inactivation. Biotechnol Prog 2017. https://doi.org/ 10.1002/btpr.2442. [14] Parker SA, Amarikwa L, Vehar K, Orozco R, Godfrey S, Coffman J, et al. Design of a novel continuous flow reactor for low pH viral inactivation. Biotechnol Bioeng 2018. https://doi.org/10.1002/bit.26497.
Acknowledgements This work has been supported by the Federal Ministry for Digital and Economic Affairs (bmwd), the Federal Ministry for Transport, Innovation and Technology (bmvit), the Styrian Business Promotion Agency SFG, the Standortagentur Tirol, Government of Lower Austria and ZIT - Technology Agency of the City of Vienna through the COMETFunding Program managed by the Austrian Research Promotion Agency FFG. The funding agencies had no influence on the conduct of this research. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.nbt.2019.10.006. References [1] Q5A(R1). International Conference on Harmonisation; guidance on viral safety evaluation of biotechnology products derived from cell lines of human or animal origin; availability–FDA. Notice. Fed Regist. 1998.
106
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al. [15] Klutz S, Lobedann M, Bramsiepe C, Schembecker G. Continuous viral inactivation at low pH value in antibody manufacturing. Chem Eng Process Process Intensif 2016. https://doi.org/10.1016/j.cep.2016.01.002. [16] Amarikwa L, Orozco R, Brown M, Coffman J. Impact of dean vortices on the integrity testing of a continuous viral inactivation reactor. Biotechnol J 2019. https:// doi.org/10.1002/biot.201700726. [17] Gritti F, Farkas T, Heng J, Guiochon G. On the relationship between band
broadening and the particle-size distribution of the packing material in liquid chromatography: theory and practice. J Chromatogr A 2011. https://doi.org/10. 1016/j.chroma.2011.09.034. [18] Knox JH, Parcher JF. Effect of column to particle diameter ratio on the dispersion of unsorbed solutes in chromatography. Anal Chem 1969. https://doi.org/10.1021/ ac60281a009.
107
Contents lists available at ScienceDirect
New BIOTECHNOLOGY journal homepage: www.elsevier.com/locate/nbt
Full length Article
A narrow residence time incubation reactor for continuous virus inactivation based on packed beds Jure Senčara, Nikolaus Hammerschmidta, Duarte L. Martinsa, Alois Jungbauera,b,* a b
Austria Centre for Industrial Biotechnology, Vienna, Austria Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria
ARTICLE INFO
ABSTRACT
Keywords: Residence time distribution Log reduction value Continuous operation Non-porous inert beads Downstream processing Plug flow
A narrow residence time distribution (RTD) is highly desirable for continuous processes where a strict incubation time must be ensured, such as continuous virus inactivation. A narrow RTD also results in faster startup and shut down phases and limits the broadening of potential disturbances in continuous processes. A packed bed reactor with non-porous inert beads was developed to achieve narrow RTDs. The performance was defined as the ratio between the onset of the cumulative RTD and the median residence time (tx%/t50%). Laboratory-scale packed columns were used to study the influence of the column parameters on the RTD. A larger column with a void volume of 0.65 L and a length of 89 cm, packed with beads in a size range of 125 to 250 μm, achieved t0.5%/ t50% > 0.93 across flow rates from 0.1 to 9.8 mL/min. The RTD was significantly narrower than the RTDs of other reactor designs, such as the Coiled Flow Inverter and Jig in a Box. The pressure drop remained under 3 kPa for all tested flow rates. Fluorescent nanoparticles (30 and 200 nm) were used to mimic viruses. These two sizes showed less than 2% difference in terms of t1%/t50% and t0.01%/t50% scores. These results indicated that viruses travelled through the column at rates independent of size. This proposal of packed beds as incubation chambers for continuous virus inactivation is simple, scalable, and can be realized as single-use devices. Due to the low pressure drop, the system can be easily integrated into a fully continuous process.
Introduction In the development of continuous integrated biopharmaceutical processes, one of the key challenges is continuous virus inactivation, because a defined, strict incubation time must be ensured [1]. This can be achieved with either a semi-continuous approach, which alternates between two batch reactors, or with a fully continuous approach, where an ideal plug flow reactor is desirable [2]. Continuous biomanufacturing is interesting, due to the potential economic benefits, such as the reduced unit operation size, the reduced footprint, and the possibility of working with disposables, even at full scale [3–5]. To render a process continuous, all unit operations need to be operated in a continuous or semi-continuous manner. A narrow residence time is required to ensure a given reaction time, but also to
limit the time for starting up and shutting down. The latter point is of particular interest when the process is interrupted. Regulatory agencies demand two orthogonal virus inactivation/removal steps for cell-culture derived biopharmaceuticals. The first step is the physical removal of viruses with nanofiltration [1]. The second requires diverse methods that depend on the process and product. The commonest ways to inactivate viruses in processing intermediates include low pH inactivation [6], solvent-detergent treatment [7], and ultraviolet C light inactivation [8]. All of these approaches work by exposing the process fluid to conditions detrimental to viruses, and incubating them in the condition for a given time. In the context of continuous operation, the batch incubation time (value) becomes a time distribution. Operated in batch, low pH inactivation and solvent-detergent
Abbreviations: λ, wavelength; t, residence time; tx%, residence time at which the normalized signal reaches x %; tbatch, batch incubation time; tmean, mean residence time; tLOD, residence time at which the signal reaches the limit of detection; FLOD, cumulative RTD at which the signal reaches the limit of detection; LOD, limit of detection; LRV, log reduction value for virus inactivation; LRV batch, required log reduction value for virus inactivation; RV, reduction value for virus inactivation; RV continuous, minimum guaranteed reduction value for a continuous virus inactivation; RV LOD, reduction value for virus inactivation for tLOD; RV preLOD, reduction value for virus inactivation for before tLOD is reached; RTD, residence time distribution; CFI, Coiled Flow Invertor; CV, column volume; JIB, Jig in a Box; PMMA, polymethylmethacrylate ⁎ Corresponding author at: Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria. E-mail address: [email protected] (A. Jungbauer). https://doi.org/10.1016/j.nbt.2019.10.006 Received 15 October 2018; Received in revised form 8 October 2019; Accepted 9 October 2019 Available online 17 October 2019 1871-6784/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
treatment are performed with exposure times of at least 30 min, but typically 60 min [9,10]. Moving towards continuous operation, a narrow residence time distribution (RTD) is of paramount importance to ensure sufficient incubation time and to prevent excess exposure of the product to harsh virus inactivation conditions. Multiple concepts for continuous or semi-continuous operation have been proposed in recent years. A semi-continuous approach towards continuous virus inactivation that employs two alternating tanks, which mimics a batch operation is currently on the market (Cadence™ Virus Inactivation system, Pall Corporation). A fully continuous virus inactivation system, based on UV-C radiation, is also currently on the market (UVivatec™, Sartorius Stedim Biotech). Recently, a fully continuous virus inactivation system that employs a size exclusion (SEC) column as the incubation reactor was reported [11]. However, those authors did not discuss the RTD or the characteristics of the reactor. Other fully continuous virus-inactivation reactor designs, with long residence times, have been reported recently in the literature, such as the Coiled Flow Inverter (CFI) [12] and the refined Jig in a Box (JIB) concepts [13,14]. Those designs were based on the induction of secondary radial flow patterns, which narrowed the RTD, compared to straight tubes. The efficiency of those reactors varies with the fluid velocity and tube diameter. The tubes employed in those types of reactors are typically very long relative to their diameter, to achieve a narrow RTD. However, that design can lead to back-pressure limitations at larger scales. A possible continuous virus inactivation setup (Fig. 1) would start with a static or dynamic inline mixer for adding the inactivation chemicals to the process fluid, and an inline sensor, such as a pH probe or a Fourier transform infrared detector, to ensure conditions remain within defined limits. The mixture then enters the incubation chamber. At the outlet of the incubation reactor, there is another inline mixer for neutralizing virus-inactivation chemicals. A 0.22 μm dead-end filter is placed before and after the inactivation chamber to control the bioburden and to remove aggregates, which could form before or during the incubation [1]. In the present study, an incubation reactor for a fully continuous virus inactivation is developed, based on well-established packed beds of non-porous inert beads. Beads with diameters between 100 μm and 1 mm were considered to achieve a narrow RTD and to avoid high backpressure and size-exclusion effects (i.e., different flow-through times for larger and smaller viruses). There was no need for custom-designed parts, because well-established chromatographic equipment was already available. In addition, the influence of design parameters on the RTD, such as column dimensions, bead size distribution, operating flow rate and the diameter of tracer nanoparticles, was investigated.
Materials and methods Equipment and buffers Chromatographic columns HR 5, HR 10, HR 16, HS 16, and XK 50 (all GE Healthcare) were packed with non-porous beads. Glass beads (Sigmund Lindner GmbH) of different diameter distributions (0.1 mm – 0.2 mm; 0.2 mm – 0.3 mm; 0.3 mm – 0.4 mm; 0.25 mm – 0.5 mm; and 1.0 mm – 1.3 mm) were used to study the influence of the bead size distribution on column performance. We tested columns with diameters in the range of 5–16 mm and heights between 3.8 and 29.3 cm. Inert beads composed of several different materials were considered for the final continuous virus inactivation application (Table 1). Based on inertness, price, sphericity, density, and robustness, polymethylmethacrylate (PMMA) beads were selected as the most appropriate. A new column was packed with PMMA beads and added to the group of glass-bead columns (Supplementary Table 1). All RTD characterization experiments were performed on an Äkta Avant™ (GE Healthcare) chromatography system. The outlet concentration was measured by the UV cell of the system. Breakthrough experiments were performed by equilibrating the column with reverse osmosis-H2O (RO-H2O), and then switching to a 2% acetone solution. The breakthrough profiles were obtained by measuring the acetone concentration at the outlet of the column, at λ =280 nm. A solvent-detergent mixture/physiological buffer system was used to investigate potential differences in column performance compared to the RO-H2O/acetone system. The solvent-detergent mixture was prepared with final concentrations of 1% Triton X-100, 0.3% Polysorbate 80, and 0.3% Tri-n-butyl phosphate, in the physiological buffer. The content of solvent-detergent chemicals was measured at λ =300 nm. To simulate viruses of various sizes traveling through the column, florescent, virus-sized nanoparticles (Micromod Partikeltechnologie GmbH, Sicastar®-greenF, plain, excitation: 485 nm, emission: 510 nm) of two different diameters (30 nm and 200 nm) were used. A short pulse (< 0.5% of the column volume) of nanoparticles (25 mg/mL) in an aqueous suspension was injected at various operating flow rates. An UltiMate™ 3000 Fluorescence Detector (Thermo Scientific) was used to detect the fluorescent nanoparticle tracers at the outlet of the column. The peak obtained was integrated, and then evaluated as a breakthrough profile. Column packing Column packing was performed with a custom-built vibrating column holder. The column was filled with water. Then the column was set to vibrate, as beads were slowly poured in at the top of the column. Column vibration continued, as they immersed in the water. Then, the column was closed and tightened, before vibration was stopped. Gravimetric testing of the packed columns showed that porosity varied by < 2% for columns packed with the same batch of beads. During 8 months of column usage, we observed no shrinkage of the bed height or decrease in column performance, based on measurements taken in periodic breakthrough experiments, including a prolonged exposure to 1 M NaOH and various aqueous buffers. Breakthrough profile characterization Peak-fronting is a primary concern in continuous virus inactivation, because the earliest volume fractions have the shortest incubation times. The narrowness of the RTD was evaluated as the ratio between the time that the concentration of the breakthrough buffer reached a certain small percentage (tx%) and the median RTD (t50%) (Fig. 2a). The closer the tx% /t50% ratio is to 1, the closer the onset of a peak is to an ideal plug flow. The Bodenstein number was obtained by fitting a breakthrough curve to the F function, as described previously [12]. In our experience, the F function does not provide a good fit to the
Fig. 1. A diagram of a potential continuous virus inactivation setup. An inline mixer continuously mixes the virus-inactivation chemicals with the process fluid. Next, the mixture passes through a 0.22 μm dead-end filter and the inline sensor (e.g., pH probe or Fourier transform infrared detector), which ensures consistent mixture composition. The mixture then flows through the incubation reactor, and after incubation, it is routed through the second inline filter. The virus inactivation properties of the process fluid are then neutralized by mixing in neutralization chemicals via a second inline mixer. 99
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Table 1 Comparison of the different types of inert beads considered for this study. Bead material (manufacturer)
Particle size distribution
Density (kg/L)
Cost
Sphericity
Glass (Sigmund Lindner GmbH) PMMA (Kisker Biotech GmbH) Ceramic (Kuhmichel Abrasiv GmbH) Polystyrene (Cospheric LLC) Polyethylene (Cospheric LLC)
broad medium medium narrow narrow
2.7 1.18 3.8 1.04 0.97
low medium low high high
medium very high high very high very high
PMMA: polymethylmethacrylate.
experimental data. Furthermore, it is not sensitive to the peak onset (Fig. 2 b). Therefore, because the peak onset is crucial for claiming a sufficient log reduction value (LRV), it was decided to use the tx%/t50% ratio as a metric for the narrowness of the RTD throughout this work. A similar concept for evaluating the narrowness of the RTD was previously employed by others [13].
performed on the column packed with PMMA beads. The tested superficial linear velocities were 3 cm/h, 9 cm/h, 30 cm/h, and 120 cm/h. Again, higher flow rates were not tested, due to the high back-pressure generated by the analytical fluorescence detector.
Investigating column parameters
Claiming the log reduction value
To investigate the influence of column parameters on the RTD, twelve columns packed with glass beads of various bead sizes and one packed with PMMA beads were characterized. Acetone breakthrough experiments were perfomed at superficial linear velocities of 50, 100, 200, and 300 cm/h. To confirm the narrowness of the RTD after scaling-up, a larger XK50 column (diameter =5 cm, height =89.0 cm, bead size = 125–250 μm) was packed and the RTD tested at superficial linear velocities that ranged from 5 cm/h (1.6 mL/min) to 30 cm/h (9.8 mL/ min). Higher flow rates were not possible, because the system pressure, generated by the UV detector, the tubing, and the valves, approached the specified pressure limit for the low-pressure XK50 column frame (< 0.3 MPa). Experiments with virus-sized fluorescent tracer nanoparticles were
Two methods have been suggested recently for achieving batchequivalency with regard to virus removal [15]. The minimum residence time approach was to set the incubation time to ensure that 99.5% of the peak eluted no earlier than the required batch incubation time (e.g. 60 min). The second approach had a more theoretical nature; it suggested selecting the incubation time to ensure that the weighted average of the LRV across the RTD was the same as the required LRV in a batch operation. Note that, in this approach, a larger portion of process fluid was allowed to have a shorter residence time than the batch incubation time. However, that approach requires a detailed analysis of virus inactivation kinetics and the exact RTD of the continuous virus inactivation reactor. It was decided to extend the minimum residence time approach by focusing on the very beginning of the RTD – even before the lower limit
Results and discussion
Fig. 2. Measurements of the narrowness of the residence time distribution (RTD) for continuous virus inactivation. (a) The narrowness is defined as the ratio between the time at the onset, when a small percent (e.g. t5%, dashed line) of the cumulative RTD has flowed through, and the median RTD (t50%, dot-dashed line). (b) The t1%/ t50% values and the Bodenstein numbers (Bn) were calculated for sample experiment breakthrough profiles with various shapes. The Bn values were similar (408 and 452) between two breakthrough profiles with similar overall shapes, but different peak fronts. In contrast, the t1%/t50% values showed better discrimination between the profiles. Thus, the t1%/t50% ratio is more sensitive to peak fronting, which is a critical feature for ensuring sufficient virus inactivation. Abbreviations: CV: column volume.
100
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
of detection (LOD) was reached. For this, a theoretical worst-case scenario was defined, based on the requirements for the batch process. A logarithmic nature for the virus inactivation kinetics is assumed. In fact, though, inactivation of virus infectivity frequently follows a biphasic curve, where a rapid initial phase is followed by a slower phase [1]. Thus, compared to a purely logarithmic function, virus inactivation typically displays much faster kinetics at the very beginning of the incubation period and slower kinetics at the end [2,6,9]. Therefore, the assumption of a logarithmic nature added an extra margin of safety to the approach. A reduction value (RV) at any time (t) was defined as the ratio between the current virus concentration (c) and the initial virus concentration (c0). The assumed virus inactivation kinetics for the batch process, with its required LRV (LRVbatch) and its required incubation time (tbatch), was described with Eq 1.
RV (t ) =
c (t ) = 10 c0
LRVbatch t tbatch
(1)
The LOD for the breakthrough profile is reached at the time (tLOD) when the normalized concentration of the species breaking through reaches a certain value (FLOD). By further assuming a single peak in the RTD, a worst-case envelope for the breakthrough profile is obtained, with a known FLOD and tLOD. The envelope of the assumed cumulative RTD rises linearly from t = 0 to tLOD, and reaches a value of 1 at tLOD (Fig. 3a). The envelope could be optimized by incorporating the t50% parameter (Fig. 3b). However, in our experience, the benefits of that optimization did not justify the additional complexity, when dealing with a very narrow RTD. The average virus RV before reaching the tLOD (RVpreLOD) was described with Eq. 2.
Fig. 4. A diagram of the worst-case theoretical scenario for a continuous virus inactivation process. We assumed a logarithmic nature of virus inactivation and a single peak in the residence time distribution. The X-axis is the required log reduction value (LRV) for an equivalent batch operation. The Y-axis is the limit of detection (LOD) for the normalized breakthrough profile (cumulative residence time distribution). The values represent the ratio of the minimum required time when the LOD is reached (tLOD) to the equivalent batch incubation time (tbatch).
Fig. 3. A worst-case assumption for a normalized breakthrough profile. (a) The assumption was based on the time point (tLOD), at which the breakthrough profile (cumulative residence time distribution) crosses the lower limit of detection (LOD). (b) The worst-case assumption can be optimized by adding information about the median residence time. In both cases, the worst-case envelope (dashed line) will always lie above the experimental data (solid line); thus, it can be safely used to approximate the normalized breakthrough profile with an additional safety margin.
101
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
RVpreLOD =
1 tLOD
tLOD 0
10
LRVbatch t tbatch dt
=
1
10
finding indicated that the input parameters were not strongly correlated over the selected range, and thus, they were suitable for a linear prediction model. The verification of the data yielded a clear trend between the measured and predicted values (Fig. 5b).
LRVbatch tLOD tbatch
LRVbatch tLOD tbatch
ln(10)
(2)
For the volume fractions after tLOD, the same RVLOD as that observed at tLOD was assumed [Eq. 3].
RVLOD = RV (tLOD ) = 10
LRVbatch tLOD tbatch
Buffer system with solvent-detergent chemicals A subset of experiments described in the previous section was repeated with the physiological buffer and the physiological buffer combined with the solvent-detergent chemicals as tracer. The results matched those from the same experiments performed with water and 2% acetone (Fig. 6). This finding justified the use of water and 2% acetone for the model buffer system. The latter buffer system was used for all subsequent experiments to achieve higher sensitivity and avoid using toxic chemicals.
(3)
The equation for the combined virus RV (RVcontinuous) was obtained by combining the reduction observed before the LOD and the reduction observed after the LOD [Eq. 4]. RVcontinuous = FLOD RVpreLOD + (1 – FLOD) RVLOD
(4)
The RVcontinuous should match the required RV for the batch operation, RVbatch [Eq. 5]. RVcontinuous = RVbatch = RV(tbatch) = 10 –LRVbatch
A column packed with PMMA beads Another analysis was performed with PMMA beads, because they were the most promising material for the intended application. Their density of 1.18 mg/mL allows for wet packing and keeps the overall weight of the column relatively low. PMMA particles were coated with a gold layer and subjected to scanning electron microscopy. Images revealed that the PMMA beads had much less surface irregularity than glass beads. In addition, no cavities or pores were detected (data not shown). The size distribution of PMMA beads was measured with a Morphology™ G3 instrument (Panalytical GmbH), and the results were within the manufacturer’s specifications (more than 95% of beads were within the specified 200–300 nm size range). The stability of the column packed with PMMA beads was also tested by prolonged flushing with 0.5 M NaOH (pH 13.7), to simulate the column regeneration/sanitization cycles, and with low pH buffer (50 mM citrate, pH 3.0), to simulate a potential low-pH inactivation mode. These chemical treatments did not alter the breakthrough profiles. The asymmetry of the column ranged from 1.1 to 1.3, depending on the superficial linear velocity (Fig. 7a). A linear log-log dependency was observed between height equivalent to a theoretical plate (HETP) and the superficial linear velocity (Fig. 7 b), consistent with an empirical model described previously [18].
(5)
By combining Eq.s 1–5, the required worst-case tLOD was calculated, based on batch process requirements (tbatch, LRVbatch) and the limit of detection (FLOD) [Eq. 6].
10
LRVbatch
= FLOD
1 10 LRVbatch
t LRVbatch t LOD
batch
tLOD tbatch
ln(10)
+ (1 – FLOD )* 10
t LRVbatch LOD tbatch
(6) The solution for Eq. 6 is presented in Fig. 4, in the form of a tLOD/ tbatch ratio depending on the LRVbatch and FLOD. The ratio approached unity as the LOD approached zero. A LOD that approached zero, combined with an ideal plug flow would cause the required mean residence time to approach tbatch. For a LOD = 0.01% and a required LRV = 5, the tLOD⁄tbatch = 1.13. In the case of a batch equivalent incubation time of 60 min, the tLOD = 68 min for a continuous virus inactivation, i.e. 68 min was the time required for the signal of the breakthrough profile to cross the lower LOD. The additional 8 min was needed to compensate for the equipment LOD, given the required LRV. A LOD of 0.01% may not be realized with existing in-line technology for any give process. Thus, an alternative approach might be needed, such as high sensitivity offline characterization. In our experience, it was possible to achieve a LOD of 0.01% with a 5% acetone solution; moreover, a LOD as low as 0.0001% was achieved in the cumulative pulse response of fluorescence nanoparticle tracers. Several potentially process-compatible tracers (that is, generally recognized as safe) were previously used for an offline characterization of JIB [16]. With the calculation that assumed the worst-case scenario for the continuous process, including experimental virus inactivation data from an established batch process, it was possible to claim an LRV at least as high as that achieved in the batch process.
Virus-sized nanoparticles The column packed with PMMA beads was also used for experiments with virus-sized fluorescent nanoparticles to investigate whether particles in the size range of viruses might travel through packed beds differently depending on their size. Fluorescent particles were used in combination with a fluorescence detector, due to the higher sensitivity compared to UV detectors. The RTDs obtained at different flow rates were evaluated in terms of the median residence time (t50%), the t0.5%⁄t50% ratio, and the t0.01%⁄t50% ratio. The difference in the results between the tracer sizes (30 nm and 200 nm bead diameters) was < 2% across the entire range of tested flow rates (Fig. 8). This finding was expected, because the bead diameters in the packed bed were much larger than the nanoparticles. An example was considered where an LRV of 3 is targeted using the column packed with PMMA beads. If a tracer with a LOD of 0.5% is used for reactor characterization, the tLOD/tmean = t0.5%/t50%. Thus we obtained a tLOD⁄tmean = t0.5%⁄t50% ≥ 0.78, based on Fig. 8, and a tLOD⁄tbatch = 1.15, based on Fig. 4. Based on Eq. 7 and a 60 min batch incubation time (tbatch =60 min), the required residence time was 89 min. To claim a LRV of 5 while using a LOD of 0.5%, the tLOD⁄tbatch value would increase to > 7, which would make the reactor impractically large. Thus, in cases that require a LRV of 5, the required LOD would be 0.01%, which gives tLOD⁄tbatch = 1.13 and tLOD⁄tmean = t0.01%⁄t50% ≥0.68 for the column packed with PMMA beads. These parameters
Influence of column parameters on RTD Twelve columns packed with glass beads and one column packed with PMMA beads were characterized at different linear velocities. The ratio t1%/t50% was correlated to column properties with a partial least square (PLS) model that had four components. The bead size distribution did not have a major influence on column performance, as found previously by others [17]. Thus, this property was excluded from the model. In the range tested, the average bead diameter had the largest influence, followed by the column length, the ratio between the column cross-section and the column length, and the flow rate (Fig. 5). The individual factors influenced the RTD (t1%⁄t50%) in the expected direction. However, no theoretical model available in the literature could describe the results quantitatively. The first PLS component alone explained 94% of the variability in the measurements (Fig. 5a). This
102
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 5. A four-component partial least square (PLS) model was employed to describe the influence of column properties on the RTD, in terms of the t1% /t50%. (a) The first component of the model (x) is compared to the corresponding linear regression model (bars). len: column length; A: column cross-section; vs: superficial linear velocity; dp: mean bead diameter. (b) Verification of the regression with the first component of the PLS model yields a trend with R² = 0.936.
time in an actual application would be expected to be closer to the batch incubation time.
tmean = tbatch *
tLOD tmean t t * = tbatch * LOD * 50% tbatch tLOD tbatch t x %
(7)
Backpressure In all columns examined, the pressure dropped below the LOD of the pressure sensor on the Äkta Avant system (< 3 kPa). Low backpressure is a desirable feature, because when multiple unit operations are directly coupled in continuous processing, the individual pressure drops add up. Fig. 9 summarizes the backpressure calculations with the Carman-Kozeny equation, assuming a porosity of 0.4 and a bead sphericity of 0.9. The equation is applicable to the examined system, because we used rigid, non-compressible particles. Fig. 9a illustrates the expected pressure drops for 250-μm particles with various residence times and column lengths. In Fig. 9b, the pressure drop across packed beds was calculated for various bead diameters and column lengths. Scaling-up To investigate the packed column performance at a larger scale, an XK 50 column was packed (diameter =5 cm, height =89.0 cm, bead size = 125–250 μm). The XK 50 column achieved a t5%/t50% value of > 0.93 across the entire tested range of linear velocities (Fig. 10). This finding indicated that this column achieved substantially narrower RTDs than the RTDs of other, smaller columns. To claim a LRV of 3, the parameters for Eq. 7 would be: tLOD⁄tmean = t0.5%⁄t50% > 0.93 and tLOD⁄tbatch = 1.15. For a batch equivalent incubation time of 60 min, the required mean residence time for the packed bed column would be 74 min. This time was significantly shorter than that attained with the much smaller column (tmean =89 min). However, to claim a LRV beyond 5 would need a lower LOD. This further emphasized the need for high-sensitivity process-compatible tracers. The CFI and JIB concepts were also implemented at larger scales; however, the Supradex 200 column [11] could only be used at a
Fig. 6. A comparison of residence time distributions (RTD) among the solventdetergent (SD) buffer system and the acetone buffer system. The RTD was evaluated in terms of the t5%/t50% ratio for various columns at various linear velocities. Individual data points correspond to two experiments performed under the same conditions, with different buffer systems. The linear trend justifies the use of the acetone buffer system as a model for the SD buffer system. Buffer systems: acetone buffer system: RO-H2O with 2% acetone as tracer; SD buffer system: physiological buffer with SD chemicals (1% Triton X100, 0.3% Polysorbate 80, and 0.3% Tri-n-butyl phosphate) as tracer.
resulted in a required mean residence time of 100 min for a batch equivalent time of 60 min. Although this residence time was significantly greater than the batch equivalent incubation time, it represented a theoretical worst-case scenario, and thus, the residence 103
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 7. Properties of a column packed with PMMA beads (d = 16 mm, h = 13.2 cm, bead size = 200–400 μm). (a) Asymmetry across superficial linear velocities. (b) A log-log relationship between superficial linear velocity and HETP. Abbreviations: PMMA: polymethylmethacrylate; HETP: height equivalent to a theoretical plate.
very low flow rate (< 1 mL/min) and it is not suitable for scaling up, due to the high backpressure generated by the SEC matrix.
performance of the CFI [12] and the JIB [13,14] is shown in Fig. 11. In terms of a narrow RTD, the smaller set of packed bed column reactors performed similarly to the CFI with much larger void volumes. Indeed, the larger packed bed column, with a void volume of 0.65 L, outperformed all reported CFI designs (void volumes up to 471 mL) and JIB designs with up to three JIB boxes in series (total void volume of 1.5 L).
Comparison with other reported designs The comparison of the packed bed performance with the
Fig. 8. Comparison of the properties of fluorescent particles with 30 nm and 200 nm diameters measured across different superficial linear velocities. (a) The median residence time, (b) the onset of the peak, in terms of t0.5% /t50% and (c) the high-sensitivity onset of the peak, in terms of t0.01% /t50%. The tx% is the residence time at which the cumulative residence time distribution reaches value of x%.
104
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 9. Expected pressure drops for columns packed with non-porous inert beads, based on the Kozeny–Carman equation. Pressure drops calculated for (a) columns of different lengths, packed with 250-μm beads, and run with various residence times; (b) columns of different lengths, with a mean residence time of 80 min, loaded with various bead diameters. For the calculations, we used a porosity value of 0.4 and sphericity value of 0.9.
Conclusion
inactivation methods, such as solvent-detergent treatments. The packed bed column attributes were correlated with the width of the RTD in a way that enabled the linear prediction of column performance. For virus-sized fluorescent particles (30 and 200 nm) it was demonstrated that the narrowness of the RTD and the median residence time did not depend on particle size (variability < 2%). This finding suggested that the passage of nanometer-sized particles through the reactor was independent of their size. The packed bed column significantly outperformed other designs,
We developed an incubation reactor based on packed beds of nonporous, inert beads for continuous virus inactivation, which offered a very narrow residence time distribution. The approach described combines existing, well-established bead chromatographic equipment and theory with a simple design and a disposable setup. The proposed system can be used for low pH inactivation, which is typically applied in monoclonal antibody downstream processes, and other virus
Fig. 10. Up-scaled processing with the XK 50 column (d =5 cm, h =89.0 cm, bead size = 125–250 μm) across different superficial linear velocities. (a) The median residence time; (b) the narrowness of the residence time distributions, in terms of the t1%/t50% ratio and (c) the t0.5%/t50% ratio, and (d) the Bodenstein number. Abbreviations: CV: column volume. 105
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al.
Fig. 11. The performance of the packed bed described in the present study compared to the performance of other designs at different scales (void volumes). The packed bed is compared to (a) the Coiled Flow Inverter design [12] and (b) the Jig in a Box design [13]. In both plots, the larger symbols translate to a narrower RTD, (a) in terms of the Bodenstein number and (b) in terms of the t0.5% /t50% ratio. All designs achieved a narrower RTD at larger scales. At similar scales, the packed bed outperforms the other two designs. The t0.5% /t50% value range in plot (b) is between 0.85 and 0.97, where the ideal plug flow would have a value of 1.
such as the JIB and CFI which had up to 2-fold larger void volumes. The results also suggested that the RTDs of all three reactor designs became narrower with increased reactor size. The design did not rely on secondary flow, such as Dean vortices; therefore, it could be readily scaled up. The scale-up could be performed by applying a wide flow regime, and thus, the process would remain well within the laminar flow region. Backpressure, an important and potentially limiting parameter in integrated processes, was particularly low for the proposed bead sizes (> 200 μm) across all the experiments.
[2] Gillespie C, Holstein M, Mullin L, Cotoni K, Tuccelli R, Caulmare J, et al. Continuous in-line virus inactivation for next generation bioprocessing. Biotechnol J 2019. https://doi.org/10.1002/biot.201700718. [3] Jungbauer A. Continuous downstream processing of biopharmaceuticals. Trends Biotechnol 2013. https://doi.org/10.1016/j.tibtech.2013.05.011. [4] Rathore AS, Agarwal H, Sharma AK, Pathak M, Muthukumar S. Continuous processing for production of biopharmaceuticals. Prep Biochem Biotechnol 2015. https://doi.org/10.1080/10826068.2014.985834. [5] Shukla AA, Gottschalk U. Single-use disposable technologies for biopharmaceutical manufacturing. Trends Biotechnol 2013. https://doi.org/10.1016/j.tibtech.2012. 10.004. [6] Mattila J, Clark M, Liu S, Pieracci J, Gervais TR, Wilson E, et al. Retrospective evaluation of Low-pH viral inactivation and viral filtration data from a multiple company collaboration. PDA J Pharm Sci Technol 2016. https://doi.org/10.5731/ pdajpst.2016.006478. [7] Horowitz B, Prince AM, Hamman J, Watklevicz C. Viral safety of solvent/detergenttreated blood products. Blood Coagul Fibrinolysis 1994. https://doi.org/10.1097/ 00001721-199412003-00006. [8] Bae JE, Jeong EK, Il Lee J, Lee JI, Kim IS, Kim JS. Evaluation of viral inactivation efficacy of a continuous flow ultraviolet-C reactor (UVivatec). Korean J Microbiol Biotechnol 2009. [9] Brorson K, Krejci S, Lee K, Hamilton E, Stein K, Xu Y. Bracketed generic inactivation of rodent retroviruses by low pH treatment for monoclonal antibodies and recombinant proteins. Biotechnol Bioeng 2003. https://doi.org/10.1002/bit.10574. [10] E3042-16. Standard practice for process step to inactivate rodent retrovirus with Triton x-100 treatment. West Conshohocken, PA: ASTM International; 2016. https://doi.org/10.1520/E3042-16. [11] Arnold L, Lee K, Rucker-Pezzini J, Lee JH. Implementation of fully integrated continuous antibody processing: effects on productivity and COGm. Biotechnol J 2019. https://doi.org/10.1002/biot.201800061. [12] Klutz S, Kurt SK, Lobedann M, Kockmann N. Narrow residence time distribution in tubular reactor concept for Reynolds number range of 10-100. Chem Eng Res Des 2015. https://doi.org/10.1016/j.cherd.2015.01.003. [13] Orozco R, Godfrey S, Coffman J, Amarikwa L, Parker S, Hernandez L, et al. Design, construction, and optimization of a novel, modular, and scalable incubation chamber for continuous viral inactivation. Biotechnol Prog 2017. https://doi.org/ 10.1002/btpr.2442. [14] Parker SA, Amarikwa L, Vehar K, Orozco R, Godfrey S, Coffman J, et al. Design of a novel continuous flow reactor for low pH viral inactivation. Biotechnol Bioeng 2018. https://doi.org/10.1002/bit.26497.
Acknowledgements This work has been supported by the Federal Ministry for Digital and Economic Affairs (bmwd), the Federal Ministry for Transport, Innovation and Technology (bmvit), the Styrian Business Promotion Agency SFG, the Standortagentur Tirol, Government of Lower Austria and ZIT - Technology Agency of the City of Vienna through the COMETFunding Program managed by the Austrian Research Promotion Agency FFG. The funding agencies had no influence on the conduct of this research. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.nbt.2019.10.006. References [1] Q5A(R1). International Conference on Harmonisation; guidance on viral safety evaluation of biotechnology products derived from cell lines of human or animal origin; availability–FDA. Notice. Fed Regist. 1998.
106
New BIOTECHNOLOGY 55 (2020) 98–107
J. Senčar, et al. [15] Klutz S, Lobedann M, Bramsiepe C, Schembecker G. Continuous viral inactivation at low pH value in antibody manufacturing. Chem Eng Process Process Intensif 2016. https://doi.org/10.1016/j.cep.2016.01.002. [16] Amarikwa L, Orozco R, Brown M, Coffman J. Impact of dean vortices on the integrity testing of a continuous viral inactivation reactor. Biotechnol J 2019. https:// doi.org/10.1002/biot.201700726. [17] Gritti F, Farkas T, Heng J, Guiochon G. On the relationship between band
broadening and the particle-size distribution of the packing material in liquid chromatography: theory and practice. J Chromatogr A 2011. https://doi.org/10. 1016/j.chroma.2011.09.034. [18] Knox JH, Parcher JF. Effect of column to particle diameter ratio on the dispersion of unsorbed solutes in chromatography. Anal Chem 1969. https://doi.org/10.1021/ ac60281a009.
107