Integration of Single Pass Tangential Flow Filtration and High Performance Countercurrent Membrane Purification for Intensification of Monoclonal Antibody Processing

ABSTRACT

There is increasing interest in the development of intensified processes that can provide significant reductions in manufacturing costs for monoclonal antibody (mAb) products. This study examines the performance of an integrated membrane process that combines single‐pass tangential flow filtration (SPTFF) with high‐performance countercurrent membrane purification (HPCMP) for the preconditioning of clarified culture fluid (CCF). The SPTFF step was operated with a single‐pass concentration factor of 20x for 24 h filtration, with no evidence of membrane fouling; the transmembrane pressure remained < 30 kPa throughout the process. HPCMP was then performed on the pre‐concentrated feed, resulting in a 60 ± 8% reduction in host cell proteins (HCPs) and more than a 30‐fold removal of DNA while maintaining 94 ± 3% mAb step yield. The HPCMP process operated continuously for 36 h with a slight decline in HCP removal in the latter stages of operation due to membrane fouling. The buffer requirement for the integrated process was only 76 L/kg mAb. These results demonstrate the potential of using an integrated SPTFF‐HPCMP process for preconditioning CCF as part of a highly intensified mAb manufacturing process with much lower cost of goods.

1. Introduction

Most monoclonal antibodies (mAbs) are currently produced by a platform downstream purification process consisting of Protein A affinity chromatography followed by a low pH hold for virus inactivation, two polishing chromatography steps, virus removal filtration, and then formulation by ultrafiltration and diafiltration (Grilo et al. 2017; Cataldo et al. 2020). The use of this type of platform can accelerate process development timelines, enabling products to be brought to market more quickly, while also reducing overall manufacturing costs by enhancing facility fit and minimizing operator training and errors.

Despite the success of the platform process, there has been renewed focus on process intensification across the industry, including the development of continuous/connected manufacturing processes, to reduce the overall cost of goods for mAb products, both for very high‐volume markets (e.g., treatments for Alzheimer's or high cholesterol) and for broader distribution of existing products in the developing world. The potential benefits of continuous processing have been discussed in a number of review papers (Zydney 2016; Gerstweiler et al. 2021; Jungbauer et al. 2024; Khanal and Lenhoff 2021).

One of the simplest approaches for process intensification is to preconcentrate the clarified cell culture fluid (CCF) before the rest of the downstream processing operations. Brinkmann and Elouafiq (2021) used single‐pass tangential flow filtration (SPTFF) for the inline concentration of CCF for several mAbs, with the preconcentration reducing the consumption of Protein A resin by as much as 80% in short fed‐batch campaigns and enabling longer column lifetimes in continuous (perfusion) processes. However, the maximum final concentration achieved by SPTFF was only 8 g/L (7x concentration factor), with some mAbs showing maximum achievable concentrations of < 2.5 g/L during single‐pass operation of a Pellicon 3 cassette. Brinkmann et al. (2018) were able to achieve much higher concentration factors using the same SPTFF system but with a CCF that was first pretreated by reducing the temperature to 2°C–12°C and the pH to 4.8 to remove DNA and HCP by flocculation; this temperature and pH adjustment would be difficult to implement in fully continuous processes. Rahane et al. (2024) were able to obtain nearly 10‐fold concentration of a CCF using SPTFF, although the maximum mAb concentration in the preconcentrated product was still < 9 g/L. The performance of the SPTFF step depended strongly on the properties of the depth filter used for the initial clarification, with the SPTFF increasing the productivity of the Protein A capture step by 1.8 to 2.6‐fold.

Although SPTFF can provide significant preconcentration of the mAb product, it typically provides minimal removal of host cell proteins (HCP). This can create significant challenges for alternative downstream processing operations like precipitation (Minervini et al. 2024; Recanati et al. 2023), crystallization (Behboudi et al. 2024; Chen et al. 2021), and aqueous two‐phase extraction (Anupa et al. 2024; Muendges et al. 2015), among others. Mohammadzadehmarandi et al. (2023) demonstrated that high performance countercurrent membrane purification (HPCMP) could provide continuous removal of HCP from CCF with a mAb yield > 90% over 48 h of continuous operation. However, long‐term operation could only be achieved by incorporating a DNA depletion step before HPCMP, and the process required significant quantities of buffer due to the relatively low mAb concentration in the feed.

The objective of this study was to investigate the potential of using a process that combines SPTFF and HPCMP processes, shown schematically in Figure 1, for preconditioning a mAb from CCF. Low‐cost hollow fiber membrane modules were used for both the SPTFF and HPCMP. SPTFF provided a volume concentration factor of 20x over 24 h of continuous operation with no evidence of fouling. HPCMP experiments performed on the preconcentrated CCF provided more than a twofold decrease in HCPs and a nearly 30‐fold reduction in DNA, with a mAb step yield of 94 ± 3%. Additionally, the HPCMP can also provide continuous buffer exchange, which would enable the removal of low MW impurities and facilitate the optimization of pH and/or conductivity before subsequent downstream operations. These results clearly demonstrate the potential of using an integrated SPTFF‐HPCMP process as a low‐cost strategy for preconditioning CCF as part of an intensified/continuous mAb manufacturing process.

Figure 1.

Schematic of integrated SPTFF and HPCMP processes for CCF concentration and HCP removal.

2. Materials and Methods

2.1. CHO Cell Culture Supernatant

Experiments were conducted using a mAb expressed in Chinese hamster ovary cells (CHO) and provided in CCF by Takeda Pharmaceuticals at a titer of approximately 4 g/L. The mAb was an immunoglobulin G2 antibody (IgG2) obtained from a fed‐batch cell culture, with a calculated pI of 7.2. The cell culture broth was harvested at a cell density of approximately 2 × 10⁷ cells/mL with cell viability ≥ 90%. The CCF was clarified by multi‐stage depth filtration, with the resulting CCF stored at –80°C. The CCF was slowly thawed at room temperature and then sterile‐filtered through 0.22 μm polyethersulfone (PES) membranes (Corning Inc., Corning, NY) before use.

2.2. SPTFF Experiments

Preconcentration experiments were performed using Revaclear 400 (Baxter, Deerfield, Il) hollow fiber polyethersulfone membranes based on results from Yehl and Zydney (2020). These modules have a total surface area of 1.8 m², fiber inner diameter of 190 μm, and a priming volume of 93 mL as reported by the manufacturer. The Revaclear 400 has a nominal molecular weight cutoff of approximately 30 kDa, with 32% rejection of myoglobin and > 99% rejection of albumin (Revaclear data sheet, Baxter Kidney Care, https://renalcareus.baxter.com/sites/g/files/ebysai3581/files/2023-05/US-RC38-230001%20V2%20-%20Revaclear%20300%20400%20and%20500%20Datasheet.pdf). The feed flow rate was controlled using a Masterflex L/S peristaltic pump (Cole‐Parmer, Vernon Hill, IL), with the permeate flow rate set by adjusting a second pump on the retentate outlet. Both pumps were fitted with Tygon long flexible life (E‐LFL) tubing (Cole‐Parmer) to minimize drift in the flow rates and to reduce particle formation. Pressures were monitored throughout each experiment using digital pressure gauges (Ashcroft, Stratford, CT) located directly on the inlet/outlet tubing. The transmembrane pressure (TMP) was evaluated as:

$$\text{TMP}=\frac{P_{\mathrm{F}}+P_{\mathrm{R}}}{2}-\frac{P_{\mathrm{P}1}+P_{\mathrm{P}2}}{2}$$ (1)

where P _F, P _R, P _P1, and P _P2 are the feed, retentate, and permeate pressures (at the module inlet and outlet), respectively. Experiments were performed with the hollow fiber module mounted in a vertical orientation, with the feed introduced through the bottom lumen‐side port while the top permeate (shell‐side) port was capped as shown in Figure 1. Before each experiment the membrane modules were flushed with a solution of 150 mM Phosphate‐Buffered Saline (PBS) at pH 7.5, taking care to remove any air bubbles from the hollow fibers and tubing.

Flux‐stepping experiments (Minervini and Zydney 2022; Minervini et al. 2024) were used to determine the maximum achievable single‐pass volume concentration factor (ratio between feed inlet and retentate exit flow rates). The experiments were conducted in total recycle mode using 2 L of CCF with both the permeate and retentate exit lines returned to the stirred feed reservoir to minimize the amount of material required during these studies. The CCF was fed into the lumen side of the hollow fiber membrane module at a feed flowrate of 60 mL/min based on the work of Yehl and Zydney (2020). The filtrate flux was increased in a stepwise fashion every 60 min, and the TMP was recorded as a function of time at each (constant) flux. Additionally, samples of the feed, retentate, and permeate were collected at the end of each step for subsequent offline analysis as discussed in Section 2.4. The filtrate flux at the onset of fouling, commonly referred to as the critical or sustainable flux (Chaubal and Zydney 2023; Field et al. 1995), was determined at the point where the TMP showed a significant increase over time during the constant flux filtration.

A long‐time filtration experiment was also performed using the same hollow fiber modules (in total recycle mode) at a fixed concentration factor of 20x for 24 h. A fresh new module was used for each experiment.

2.3. HPCMP Experiments

HPCMP experiments were performed using small hollow fiber modules manufactured by 3 M (Saint Paul, MN) and described in more detail by Mohammadzadehmarandi et al. (2023). The modules consist of 26 hollow fibers with an effective filtration length of 23 cm, giving a total surface area of 40 cm². A new module was used for each experiment.

The 20x concentrated CCF obtained from the SPTFF was sterile filtered through a 0.22 μm PES syringe filter (Cytiva, Marlborough, MA) and fed into the lumen‐side of the membrane module using a Masterflex peristaltic pump (Cole‐Parmer) at a flow rate of 0.005 mL/min. The draw solution, consisting of a 150 mM PBS buffer at pH 7.5, was introduced through the shell‐side port at the far end from the feed inlet (i.e., countercurrent to the concentrated CCF solution as shown in Figure 1) at a flowrate of 0.02 mL/min; the use of a draw solution to feed solution flow ratio of α = 4 is consistent with previous experimental (Mohammadzadehmarandi et al. 2023) and theoretical (Yehl and Zydney 2022) work. The draw solution flow rates, both into and out of the module, were controlled using a single pump (Masterflex, Cole‐Parmer) fitted with two pump heads to maintain zero net ultrafiltration.

The hollow fiber modules were initially flushed with DI water, taking care to remove any air bubbles. The modules were then flushed with PBS before switching the feed to the preconcentrated CCF solution. The extent of ultrafiltration was monitored by measuring the change in weight of the feed and draw solution reservoirs (placed on separate digital balances) over the course of the experiment. The retentate and draw solution exit flow rates were determined by timed collection. Samples were collected periodically from the feed reservoir, draw solution outlet, and retentate solution outlet for offline analysis.

The use of these hollow fiber modules for buffer exchange was examined by performing a separate experiment using a draw solution spiked with 1 g/L of vitamin B₁₂ (Caymen Chemicals, Ann Arbor, MI) as a model solute (1355 Da). The concentration of vitamin B₁₂ in the CCF retentate outlet was evaluated by UV absorbance at 550 nm using an Infinite m200 Pro microplate reader (Tecan Trading AG, Switzerland). An appropriate calibration curve was constructed using preconcentrated CCF spiked with known concentrations of vitamin B₁₂ to ensure accurate quantification of the vitamin B₁₂ in the presence of mAb and HCP.

2.4. Analytics

mAb concentrations were evaluated by analytical Protein A affinity chromatography using a 0.1 mL POROS A column with 20 µm particles (Thermo Fisher Scientific, Waltham, MA). Data were obtained using a Vanquish UHPLC, with the protein concentration determined by UV absorbance at 280 nm evaluated by a photodiode array detector (Thermo Fisher Scientific). HCP concentrations were evaluated using a 3rd generation CHO ELISA kit (Cygnus Technologies, Southport, NC). DNA concentrations were evaluated using a Qubit 1X dsDNA BR fluorescence assay (Thermo Fisher Scientific).

Additional characterization of the CCF was performed by size exclusion chromatography (SEC) using an Agilent 1260 Infinity II HPLC system (Agilent Scientific instruments, Santa Clara, CA) equipped with an Agilent Bio SEC‐3 column with 3 µm particles having 100 Å pore size (Agilent Scientific instruments). The SEC was done at a constant flow rate of 0.35 mL/min with the UV‐Vis detector at 215 nm. Limited samples were also examined by SDS‐PAGE using NuPAGE Novex Bis‐Tris mini gel with NuPAGE MES SDS Running Buffer and Imperial Protein Stain (all from Thermo Fisher Scientific).

3. Results and Discussion

3.1. Concentration of CCF Using SPTFF

The appropriate operating conditions for continuous single pass TFF were determined by flux‐stepping experiments; the CCF feed flow rate was set at 60 mL/min based on the results reported by Yehl and Zydney (2020) for the concentration of human serum immunoglobulin G. Typical data are shown in Figure 2. The TMP (from Equation 1) and the feed‐side pressure drop (∆P = P _F ‐ P _R) are both shown as a function of time, with each 60‐min interval on the x‐axis corresponding to operation at a constant single‐pass volume concentration factor (labeled directly above the TMP data). Thus, the initial cycle at 3x was performed with an inlet feed flow rate of 60 mL/min and a retentate (exit) flow rate of 20 mL/min, corresponding to a constant permeate flux of 1.3 L/m²/h. Although the flux‐stepping experiments were performed in total recycle mode, the mAb concentration in the feed reservoir decreased over the course of the experiment due to the accumulation of the highly concentrated mAb in the module and retentate recycle line. For example, the initial mAb concentration was 4.0 g/L but this decreased to approximately 3 g/L for SPTFF at a 10x concentration factor and then to 2.2 g/L after 360 min. The mAb concentration stabilized within the first few minutes of operation at each volume concentration factor.

Figure 2.

Flux‐stepping experiment performed with CCF at a feed flow rate of 60 mL/min using 1.8 m² hollow fiber membrane module. Volume concentration factor is reported for each 60‐min step.

The TMP often exhibited a slight increase at the start of each constant flux filtration; this is most pronounced for the initial cycle at 3x and is likely due to transients associated with the fairly large hold‐up volume within the hollow fiber module. The rapid equilibration is consistent with the very short residence times (< 2 min) in the SPTFF module. In each case, the TMP stabilized during the cycle up to a volume concentration factor of 26x (with a TMP gradient of < 1 kPa/h). However, a further increase in the concentration factor to 40x resulted in a small but significant increase in the TMP (from 30 to 33 kPa over 30 min), corresponding to a TMP gradient of approximately 6 kPa/h. This increase in TMP is likely due to the accumulation of foulants on or within the membrane and suggests that it would not be possible to perform the SPTFF under these conditions for an extended continuous process without frequent membrane replacement.

The axial pressure drop through the module remained nearly constant at a value of around 1 kPa throughout the flux‐stepping experiment; the slight decrease in ∆P at very high concentration factors is due to the large reduction in feed flow rate in the latter half of the module. The absence of any increase in ∆P suggests that fiber blockage is negligible under the conditions of these experiments (Minervini et al. 2022, 2024). The mAb step yield, as determined from analytical Protein A chromatography, was > 99% for all operating conditions below the critical flux (or critical concentration factor), reflecting the absence of any significant fouling under these conditions (Bacchin et al. 2006). Permeate samples showed no evidence of mAb transmission through the Revaclear membrane (mAb concentrations less than 0.01 g/L even at the highest concentration factor). Repeat measurements performed on a separate CCF (with a different mAb) provided stable operation with a concentration factor of 30x as shown in the Supporting Information.

To verify that the conditions identified in the flux‐stepping experiments will enable the long‐term stable operation of an SPTFF module, a 24‐h filtration experiment was performed using 2 L of CCF at a feed concentration of 2.8 g/L and a feed flow rate of 60 mL/min. The volumetric concentration factor was maintained at 20x (based on the results in Figure 2) by adjusting the retentate exit flow rate to 3 mL/min. The TMP and ΔP profiles are shown as a function of time in Figure 3. There was no evidence of any fouling, with both the TMP = 26 kPa and Δp = 0.8 kPa remaining essentially constant throughout the entire experiment. However, it is important to note that the recirculation of the CCF exposes the membrane to much lower amounts of impurities; data obtained in true single‐pass operation (requiring much larger quantities of feed) will be needed to demonstrate the effectiveness of this process for long‐time CCF concentration. The outlet mAb concentration remained steady at 56 ± 0.5 g/L over the 24‐h filtration, resulting in a step yield of 100 ± 1% as determined by Protein A chromatography. SEC analysis of the final product solution showed no evidence of aggregate formation, with the concentration of high MW species (eluting before the mAb) of < 5%, which was essentially identical to that in the original CCF. Note that this likely includes both mAb aggregates as well as other large components present in the CCF, e.g., DNA and chromatin. The 24‐h filtration examined in Figure 3 would have been able to process nearly 90 L of CCF, corresponding to 0.25 kg of mAb, using a single hollow fiber membrane module operated in single‐pass mode without recirculation of the feed or the need for any intermediate cleaning and/or regeneration.

Figure 3.

24‐h filtration experiment performed with CCF at a feed flow rate of 60 mL/min using 1.8 m² hollow fiber membrane module. Data were obtained in total recycle mode utilizing 2 L of CCF feed at a volume concentration factor of 20x.

3.2. HPCMP for mAb Purification

The SPTFF process examined in Figure 3 provides a significant degree of mAb concentration, but the HCP concentration was reduced by < 15% due to the high HCP retention by the hollow fiber module. To obtain a significant reduction in HCP, we employed High Performance Countercurrent Membrane Purification (HPCMP) using a somewhat larger pore size membrane (nominal molecular weight cutoff of approximately 50 kDa). These membranes were operated with diffusive transport to enhance the selectivity for HCP removal compared to what can be achieved during pressure‐driven filtration (Mohammadzadehmarandi et al. 2023). The feed to the HPCMP module was the 20‐fold concentrated CCF solution with a mAb concentration of ≈56 g/L and an HCP content of slightly more than 100,000 ppm (as determined by CHO ELISA). The system was operated in true single‐pass mode (no recycle) at a feed flow rate of 0.005 mL/min, with this value based on results reported by Mohammadzadehmarandi et al. (2023). The hold‐up volume in the system was approximately 2 mL; thus, it took more than 8 h to flush the buffer out of the system. The mAb concentration in the product outflow was undetectable for the first 6 h and then slowly increased with time due to dilution effects in combination with dispersion within the system.

Figure 4 shows results obtained after complete wash‐out of the initial buffer, with data shown for the mAb step yield (top panel) and the HCP concentration in ppm relative to the mAb (bottom panel), with the step yield defined as:

$$\mathrm{Y}(\%)=\frac{{q_{\mathrm{R}}C}_{\mathrm{R}}}{q_{\mathit{F}}{C}_{\mathrm{F}}}\ast 100$$ (2)

where q _F and q _R are the lumen‐side feed and retentate volumetric flowrates, respectively, while C _F and C _R are the mAb concentrations in the feed and retentate. The net ultrafiltration rate, determined by measurements of q _R and q _F, was less than 5%. The mAb concentration in the retentate exit remained nearly constant at 52.5 ± 1.4 g/L, corresponding to a mAb step yield of 94 ± 3%. Most of the mAb loss was due to diffusion into the draw solution, although at long operating times the membrane became more retentive (due to protein fouling) with the mAb concentration in the draw solution outlet decreasing to < 0.1 g/L. This was also consistent with the nearly 10‐fold reduction in hydraulic permeability for the membrane determined at the end of the 36‐h experiment (evaluated from data for the buffer flux as a function of the transmembrane pressure) compared to that of the clean membrane. The high degree of fouling observed in the HPCMP process is likely due to the high level of DNA in the preconcentrated CCF, approximately 2,000,000 ppb, which has previously been shown to be the dominant foulant during HPCMP operation (Mohammadzadehmarandi et al. 2023).

Figure 4.

mAb step yield (top) and HCP concentration (bottom) during a 36‐h HPCMP experiment performed with 20x preconcentrated CCF. Data obtained at a feed flow rate of 0.005 mL/min and a draw solution flowrate of 0.02 mL/min. Horizontal error bars represent the time intervals for sample collection, while vertical error bars indicate the confidence limits based on several repeat measurements (at least three measurements per time point). The time point at t = 0 corresponds to the introduction of the feed; data are shown for time points after wash‐out of the hold‐up volume within the system.

The HCP concentration in the retentate product decreased from 105,000 ± 2000 ppm in the feed to 44,000 ± 5000 ppm for the samples obtained during the initial phase of the HPCMP run; the HCP concentration increased to 58,000 ± 4000 ppm for the final retentate sample obtained after 36 h of continuous operation. The reduction in HCP clearance is again due to fouling, with the membrane becoming somewhat more retentive to HCP over the course of the 36 h experiment. Since the HPCMP was run at approximately zero ultrafiltration, the TMP was essentially zero throughout the run. The feed‐side pressure drop for this experiment remained nearly constant over the 36 h with ∆p < 0.1 kPa due to the very low feed flowrate.

The HPCMP also provided significant DNA removal, with the DNA concentration decreasing from around 10,000,000 ppb in the feed to only 300,000 ± 20,000 ppb in the retentate product (concentrations relative to the mAb). This 30‐fold reduction in DNA is due to both DNA removal, with DNA detected in the draw solution from the HPCMP, in addition to DNA adsorption/deposition on the membrane surface.

3.3. Buffer Exchange

In addition to providing significant HCP and small solute removal, the HPCMP process can also provide true buffer exchange, with the solution characteristics of the preconcentrated CCF adjusted based on the composition of the draw solution. This was examined experimentally by performing a separate experiment in which the draw solution in the HPCMP process was spiked with vitamin B₁₂ at a concentration of 1 g/L, with the results presented in Figure 5. The vitamin B₁₂ concentration in the retentate (product) outlet during the 36‐h HPCMP process remained nearly constant at a value of 1.1 ± 0.1 g/L, corresponding to > 99% buffer exchange. This approach could easily be used to adjust the pH, conductivity, and/or excipient concentration before subsequent downstream processing steps. Previous studies of similar hollow fiber membranes showed stable operation during blood dialysis for more than 96 h (Pannu and Gibney 2020), suggesting that it may be possible to operate these modules for even longer periods of time.

Figure 5.

Vitamin B₁₂ concentration in the retentate outlet during a 36‐h HPCMP experiment performed with 20x preconcentrated CCF. Data obtained at a feed flow rate of 0.005 mL/min and a draw solution flowrate of 0.02 mL/min, with the draw solution containing 1 g/L of vitamin B₁₂. Horizontal error bars represent the time intervals for sample collection, while vertical error bars indicate the confidence limits based on several repeat measurements.

3.4. Integrated Process Performance

The performance of the SPTFF‐HPCMP combination was examined in more detail by SEC and SDS‐PAGE, with results summarized in Figures 6 and 7, respectively. The original feed material (blue curve in Figure 6) shows a strong peak for the mAb at 9.2 min, with a significant concentration of low MW species with elution times between 12 and 17 min. Note that ribonuclease A (MW = 13.7 kDa) has an elution time of 11.9 min, so the peaks at elution times beyond 12 min are relatively low MW species, including many media components (e.g., amino acids). The bottom panel of Figure 6 shows an enlarged image of the SEC chromatogram in the region occupied by the HCP. SPTFF was very effective at removing species with elution times > 14 min but provided minimal removal of HCP with elution times < 13 min. However, the combined SPTFF‐HPCMP process provided complete removal of all peaks with retention times > 13.5 min and approximately 60% reduction in total HCP based on the integrated peak area between 9.8 and 13 min.

Figure 6.

SEC chromatograms (top) of the initial feed, the 20x concentrated material obtained by SPTFF, and the retentate product obtained after HPCMP (after 12 h of operation). Bottom panel shows an expanded view of the region containing HCP. The UV absorbance has in each case been normalized by the mAb concentration in that sample as determined by analytical Protein A chromatography.

Figure 7.

Coomassie blue SDS‐PAGE gel image. M is the protein ladder. 1 = original feed (before the flux stepping experiment). 2 = SPTFF feed (during the 20x concentration). 3 = SPTFF permeate. 4 = HPCMP feed (dilution 1:20). 5 = HPCMP draw solution after 12 h. 6 = HPCMP product after 12 h (dilution 1:20). 7 = HPCMP draw solution after 36 h. 8 = HPCMP product after 36 h (dilution 1:20). Loading volume for 1 is 3 µL; for 2, 4, 6, and 8 is 5 µL; and for 3, 5, and 7 is 20 µL.

The SDS‐PAGE results (non‐denaturing conditions) are presented in Figure 7. The original CCF feed (Lanes 1 and 2) shows a distinct band at 150 kDa, corresponding to the mAb, along with a wide range of HCPs with MW spanning from 10 to 100 kDa. The most prominent impurity‐related bands are observed at MWs of approximately 25, 70, and 100 kDa. The permeate sample from the SPTFF (Lane 3) exhibits faint bands below 25 kDa, in agreement with the limited purification achieved throughout the pressure‐driven filtration. The retentate obtained from the early stage of the HPCMP process (lane 6) shows a significant reduction in HCP with MW < 25 kDa, consistent with the greater band intensity within this range for the corresponding sample from the draw solution in Lane 5). Note that the sample from the draw solution obtained during the initial phase of the process showed a significant amount of mAb but this decreased in the later stages of the process (Lanes 7 and 8).

The performance of the SPTFF‐HPCMP process is summarized in Table 1. The mAb concentration was increased from 2.8 to 52.5 g/L with an overall yield of 94 ± 3%, mostly due to a small yield loss during HPCMP. The HCP concentration was reduced from 120,000 ± 2,000 to 49,000 ± 10,000 ppm (based on ELISA), corresponding to a log reduction value of LRV = 0.4 ± 0.1, with the large majority of that occurring in the HPCMP process. The overall small molecule clearance (determined by the SEC peak area for elution times from 14 to 17 min) was more than 99%. The DNA content was also decreased significantly, from an initial level of around 10,000,000 to 300,000 ± 20,000 ppb, corresponding to LRV = 1.5 ± 0.1. This unexpected reduction in DNA is likely attributed to both DNA removal and adsorption/deposition on the membrane surface during both the SPTFF and HPCMP processes.

Table 1.

mAb Concentration (determined by analytical Protein A), HCP concentration (determined by CHO‐ELISA), and DNA concentration (determined by Qubit assay) during SPTFF‐HPCMP process. HPCMP data are averaged over the 36‐h operation.

	mAb Concentration (g/L)	HCP (ppm)	DNA (ppb)
CCF	2.8	120,000 ± 2000	10,000,000 ± 800,000
SPTFF product	56 ± 0.5	105,000 ± 2000	2,000,000 ± 200,000
HPCMP product	52.5 ± 1.4	49,000 ± 10,000	300,000 ± 20,000

The operational parameters for the integrated continuous SPTFF‐HPCMP process examined in this study are summarized in Table 2. The calculations were performed based on the outflow of a 500 L perfusion bioreactor operating with one volume exchange per day and 2.8 g/L mAb titer (neglecting the effects of any cell bleed). The SPTFF process for the 20‐fold concentration of CCF using a feed flux of 12 L/m²/h requires approximately 10 m² of membrane surface area. This corresponds to six hollow fiber cartridges in parallel, each with an individual cost of around $30/module. This yields a cost of approximatively $0.13/g mAb for the membrane modules assuming 24‐h of continuous operation; this would be reduced to < $0.02/g mAb if the SPTFF process could be run for ≥ 7 days. Further reductions in cost could be achieved by cleaning and re‐using the membrane modules.

Table 2.

Design calculations for a continuous integrated SPTFF‐HPCMP process for preconditioning CCF assuming a 500 L perfusion bioreactor operated with one volume exchange per day, a mAb titer of 2.8 g/L, and assuming 7 day operating time for the membrane modules. The countercurrent dialysis module would be used in place of HPCMP for buffer exchange. Details are provided in the text.

	SPTFF (20x)	HPCMP	Dialysis
Feed flow rate (L/h)	21	1	1
Feed flux (L/m2/h)	12	0.075	0.4
Feed concentration (g mAb/L)	2.8	56	56
Membrane area (m2)	10	14	3
Membrane capacity (kg/m2)	1	0.7	3
Buffer required (L/kg mAb)	—	76	38

The HPCMP process was performed at a feed flux of 0.075 L/m²/h. The 20x concentrated feed from SPTFF process would have a flow rate of 1.0 L/h; thus, the HPCMP process would require around 14 m² total membrane area and would use 76 L of buffer/kg of mAb, which is only a small fraction of that for a typical mAb process, which can be as much as 25,000 L/kg of buffer (Budzinski et al. 2019).

Although the low feed flux in the HPCMP process is needed to achieve high levels of HCP removal, high levels of buffer exchange can be accomplished using much higher feed flow rates (shorter residence times). The buffer exchange efficiency for the countercurrent contactor can be evaluated as (Yehl and Zydney 2022):

$$\frac{C_R}{C_{D0}}=\frac{\alpha \ast \left(\text{exp}\left(\frac{k_0\ast A}{q_F}\ast \left(1-\frac{1}{\alpha }\right)\right)-1\right)}{\alpha \ast \text{exp}\left(\frac{k_0\ast A}{q_F}\ast \left(1-\frac{1}{\alpha }\right)\right)-1}$$ (3)

where C _R and C _D0 are the solution concentrations in the retentate exit and draw solution inlet, respectively, α is the ratio of the draw to feed flowrate, k ₀ is the overall solute mass transfer coefficient, and A is the membrane surface area. Previous results showed a vitamin B₁₂ mass transfer coefficient of k ₀ = 3.5 ∗ 10⁻⁷m/s (Yehl et al. 2019); this value is also consistent with the data in Figure 5. The overall mass transfer coefficients for the HCP are considerably smaller, reflecting the smaller diffusivities and the greater hindrance to diffusion through the membrane pores. Based on Equation (3), an HPCMP process with buffer exchange efficiency of 95% could be achieved using a feed flux of q _F/A = 0.4 L/m²/h. Note that the mass transfer coefficients for typical buffer components (e.g., NaCl or histidine) would be several‐fold larger than that of the vitamin B₁₂, enabling operation of the HPCMP process at feed flux well above 1 L/m²/h and/or at smaller draw solution flow rates.

4. Conclusions

The results presented in this study clearly demonstrate the feasibility of employing an integrated SPTFF‐HPCMP process for the pretreatment of CCF before subsequent downstream operations. The SPTFF module was operated continuously (without intermediate cleaning/regeneration) for 24 h at a volume concentration factor of 20x, yielding a mAb concentration of 56 g/L. This is well beyond the < 10 g/L obtained in previous work by Brinkmann and Elouafiq (2021), Brinkmann et al. (2018), and Rahane et al. (2024). HPCMP experiments performed using pre‐concentrated CCF with a draw to feed flowrate ratio of α = 4 provided a mAb step yield of 94 ± 3%, an HCP reduction of around 60 ± 8% (determined by CHO ELISA), and more than a 30‐fold reduction in DNA (determined by Qubit assay) while simultaneously performing nearly complete buffer exchange; this is the first successful demonstration of HPCMP with a preconcentrated CCF. Although the HPCMP exhibited limited clearance of high MW HCPs, it effectively removed nearly all impurities with MW below 25 kDa, as confirmed by both SEC and SDS‐PAGE analyses. Future studies using proteomic analysis, e.g., by LC‐MS, could be used to identify those proteins that are, and aren't, removed by HPCMP. Note that the HCP removal slightly decreased throughout the 36‐h continuous process due to membrane fouling, which was likely associated with the high DNA content in the concentrated CCF.

The integrated SPTFF‐HPCMP process would be particularly attractive before subsequent downstream operations that have greater efficiency when operated at high mAb concentrations. For example, the 20‐fold increase in mAb concentration would dramatically reduce the buffer required for washing the solid phase in continuous precipitation/crystallization, while at the same time reducing the mass of excipients (per unit mass of mAb) needed for the process due to the shift in protein equilibrium. For example, Mergy et al. (2024) have obtained preliminary data showing that a downstream process incorporating mAb precipitation plus two flow‐through chromatography steps (without any affinity step) could provide similar HCP and DNA removal as a platform process employing Protein A bind‐and‐elute chromatography. In addition, the buffer exchange provided by HPCMP ensures the elimination of small interfering components, while providing any required pH and conductivity adjustment and enabling standardization of precipitation/crystallization conditions with reduced concentration of precipitants (Zhu et al. 2025). This integrated pretreatment would also significantly improve the performance of the flow‐through chromatography steps, both by reducing the process volume and lowering the amount of HCP/DNA impurities. Future studies will be required to evaluate the overall economics/performance of downstream processes incorporating CCF pretreatment by the combination of SPTFF + HPCMP, and it will also be important to explore the integration of this technology with a perfusion bioreactor (with appropriate cell recycle and clarification) as part of a truly continuous/interconnected mAb production process.

Author Contributions

Mirko Minervini: data curation, formal analysis, investigation, and writing – original draft. Aylin Mohammadzadehmarandi: data curation, investigation, formal analysis, and writing – review and editing. Ali Behboudi: investigation, formal analysis, and writing – review and editing. Sheldon F. Oppenheim: resources, conceptualization, and writing – review and editing. Andrew L. Zydney: conceptualization, funding, supervision, and writing – review and editing.

Disclosure

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Information.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.