Use of LC‐MS to characterize host cell protein removal during depth filtration

Abstract

The removal of host cell proteins (HCPs) is crucial in biopharmaceutical production, as residual impurities can impact product safety and efficacy. While a number of studies have demonstrated that depth filtration can provide significant HCP removal, there is little information on its effectiveness in removing specific HCPs. This study examines the application of liquid chromatography‐mass spectrometry (LC‐MS) to track HCP removal during depth filtration, providing a detailed analysis of HCP behavior with two commercial depth filters. Our findings reveal significant variability in HCP breakthrough behavior, with transmission patterns showing minimal correlation with either the protein isoelectric point or hydrophobicity, highlighting the unique behavior of individual HCPs. Both the X0SP and X0HC depth filters achieved almost complete removal of Lipoprotein Lipase, and the X0SP filter also effectively removed Lysosomal Acid Lipase (LAL), both known to degrade polysorbate in monoclonal antibody formulations. However, neither filter provided significant removal of Alpha‐enolase, Carboxypeptidase D, Glutathione S‐transferase, or Phospholipase B‐like 2. The X0SP filter showed equal or better removal for 18 out of 20 problematic HCPs, with greater HCP removal seen at lower conductivity. This work provides a detailed framework for understanding and optimizing depth filtration processes, offering insights into the effectiveness of depth filters for removal of problematic HCPs.

1. INTRODUCTION

The removal of host cell proteins (HCP) is a critical component of the downstream process in the production of biopharmaceuticals. Most monoclonal antibody (mAb) processes are designed to reduce HCP levels to below 100 ppm (100 ng/mg product), but the presence of specific HCP can affect the safety and efficacy of the final product at much lower concentrations due to their immunogenicity or biological activity. ¹ Thus, the accepted level of HCP in the final product must be determined by an appropriate risk assessment analysis. ²

Although the bulk of the HCP removal in a typical mAb process occurs by chromatography, including both Protein A and ion exchange, there is a growing realization that depth filtration used for initial clarification can also provide a significant reduction in HCP levels. ³ Yigzaw et al. ³ provided one of the first demonstrations that depth filtration can reduce HCP concentrations, in this case resulting in a significant reduction in turbidity in the eluate from the Protein A affinity step. Subsequent studies by Khanal et al. ⁴ have identified the relative contributions from the different components of the depth filter (e.g., filter aid, binder, and cellulose), while Nejatishahidein et al. ⁵ and Chu et al. ⁶ evaluated the effects of protein isoelectric point and hydrophobicity on protein binding to different depth filters using a series of model proteins. Nguyen et al. ⁷ evaluated the effectiveness of depth filtration on HCP removal immediately after Protein A affinity chromatography, with the fully synthetic X0SP filter reducing HCP levels by more than 50% (to below 100 ppm) for 4 out of 7 mAbs/fusion proteins.

Although these studies have provided important insights into the effectiveness of depth filtration for HCP removal, they have lacked the specificity to evaluate the behavior of individual HCP. Recent advances in liquid chromatography‐mass spectrometry (LC‐MS) have provided an invaluable tool that can identify and quantify specific HCP, often down to parts‐per‐million (ppm) levels. ⁸ , ⁹ LC‐MS has been used successfully to examine HCP levels after Protein A ¹⁰ and ion exchange chromatography. ¹¹ Gilgunn et al. ¹² used LC‐MS to examine HCP removal through a process employing the Emphaze™ AEX Hybrid Purifier during clarification and Protein A chromatography. The addition of the Emphaze™ provided high levels of HCP removal, including the removal of several problematic HCP (e.g., 78 kD glucose‐regulated protein, nidogen‐1, heatshock protein, actin, serine protease HTRA1, and matrix metalloproteinase‐19). However, no quantitative information was provided on the HCP removal by the depth filter alone.

The objective of this study was to examine the potential of using LC‐MS to characterize HCP removal during depth filtration, including quantitative analysis of the breakthrough behavior of individual HCP generated from a null CHO cell line. HCP removal varied dramatically for individual proteins, but the binding behavior showed limited correlation with either the protein isoelectric point or the protein hydrophobicity (as determined by the Grand Average of Hydropathy or GRAVY number ¹³ ). The results not only provide important insights into the HCP removal characteristics of different depth filters, but they also provide a framework for future studies aimed at optimizing the use of depth filtration in the development of improved downstream processes for biopharmaceutical manufacturing.

2. MATERIALS AND METHODS

2.1. Host cell protein preparation

All experiments were performed with a null CHO cell line (no expression of any recombinant proteins) to eliminate interference in the LC‐MS analysis. Chinese Hamster Ovary cells (CHO‐S cells, Catalog Number R80007, Thermo Fisher) were thawed from previously frozen aliquots and seeded in shaker flasks (125 mL Corning Erlenmeyer flasks without baffles) at an initial concentration of 1–5 × 10⁵ cells/mL. Cells were expanded in Freestyle™ CHO Expression media (Catalog Number 12651014) with 8 mM L‐Glutamine (Catalog Number 25030081), all from Thermo Fisher, at 37°C using 8% CO₂, 95% humidity, and 110–115 rpm agitation in a New Brunswick S41i shaker incubator (Eppendorf, Germany) to achieve a concentration of 2–4 × 10⁶ cells/mL before use or subculture. Cell viability was maintained at >96%.

The cells were centrifuged at 450g for 5 min at 4°C to obtain a pellet with a cell density of approximately 200 × 10⁶/mL. The cell pellet was stored at −20°C until the day of the experiment. The cells were resuspended using 1× PBS (in a 1:1 ratio by volume) and then lysed by sonication for 30 min on a floating foam rack using a 40 kHz Heated Ultrasonic Cleaner (Thermo Fisher Scientific, Waltham, MA) to increase the HCP concentration. Note that previous work by Tait et al. ¹⁴ showed that differences in HCP abundance between a null and antibody‐producing CHO cell line were due primarily to cell lysis associated with the loss of cell viability due to over‐expression of the antibody; thus, the cell lysis step used in this work should provide a better model for the HCP profile in a mAb‐producing cell line. Limited experiments were performed with HCP mixed with a purified monoclonal antibody and with cell pellets resuspended in 0.3× PBS with an ionic strength of approximately 45 mM to study the effect of conductivity on HCP removal. The cells were then fully thawed using a room temperature water bath, thoroughly mixed by pipette aspiration, and then sonicated for another 30 min. The lysed cells were centrifuged at 8000g for 20 min and then filtered through a 0.2 μm polyethersulfone (PES) sterile filter (Thermo Fisher Scientific, Waltham, MA), with the permeate used as the CHO host cell protein (HCP) feed solution.

2.2. Depth filtration

Depth filtration experiments were performed using X0SP and X0HC fine grade depth filters in Millistak+ ProPod modules from MilliporeSigma. The X0SP filter has a synthetic silica‐based filter aid supported by polyacrylic fibers, while the X0HC has diatomaceous earth supported by cellulose fibers. The depth filters were initially flushed with deionized water, followed by PBS, and then challenged with the HCP solution at a filtrate flux of 150 L/m²/h, which was set using a Masterflux pump (Cole‐Parmer). Additional details on the depth filter media and the experimental procedures are provided in an earlier publication. ⁶ Permeate samples were collected for offline analysis of the total host cell protein concentration by Bradford assay or Qubit™ fluorometric quantification assay using albumin standards. ¹⁵ Appropriate quantities of each permeate sample, with approximately the same total mass of HCP, were then lyophilized using a benchtop freeze dryer (Labconco, Kansas City, MO), providing a stable matrix that could be used to store the proteins until subsequent analysis.

2.3. Liquid chromatography‐Mass spectrometry (LC‐MS)

The lyophilized samples were mixed with 50 μL of a solution containing 5% SDS and 50 mM Triethylammonium bicarbonate (TEAB) buffer at pH 7.1. Tris(2‐carboxyethyl) phosphine (TCEP) (Thermo Fisher) was added to a concentration of 10 mM, and the mixture was incubated at 65°C for 30 min. ¹⁶ The solution was cooled to room temperature; 1 μL of a 0.5 M iodoacetamide acid solution was added, and the mixture allowed to react for 30 min in the dark. The reaction was quenched by adding 2.7 μL of a 12% phosphoric acid solution. The protein solution was then mixed with 165 μL of binding buffer (90% Methanol, 100 mM TEAB; pH 8.5) and purified using an S‐Trap spin column (ProtiFi, Farmingdale NY) on a bench top centrifuge (60 s spin at 1000 g). The spin column was washed twice with 150 μL of 50% methyl‐tert butyl ether with 50% methanol and then twice with 150 μL of binding buffer.

Protein digestion was performed by addition of 600 ng of trypsin (Promega, #V5280, Madison, WI) with the resulting mixture incubated overnight at 65°C. The protein digest was eluted twice with 75 μL of a solution containing 50% acetonitrile and 0.1% formic acid in water. The peptide concentrations were evaluated by a fluorometric peptide assay (Thermo Fisher, 23,290). Approximately 230 ng of the peptide sample was then dried in a speed vacuum (room temperature, 1.5 h) to remove solvents. The dried peptide samples were resuspended in 10 μL of 2% acetonitrile, 0.1% formic acid, and 97.9% water and then aliquoted into autosampler vials.

The peptide mixtures were analyzed by liquid chromatography–tandem mass spectrometry (NanoLC‐MS/MS) using an UltiMate 3000 RSLCnano LC chromatography system (Dionex) coupled to a Thermo Orbitrap Fusion mass spectrometer (Thermo Fisher) with a nanospray ion source in the core facility at the University of Texas Medical Branch. Samples were processed through an Acclaim PepMap trap column (100 μm × 2 cm, 5 μm; Thermo Fisher, 164,750) followed by an Acclaim PepMap analytical column (75 μm × 2 cm, 3 μm; Thermo Fisher, 164,569). Both columns were initially equilibrated in 98% solvent A (0.1% formic acid in water) and 2% solvent B (0.1% formic acid in acetonitrile). Gradient elution was performed at 300 nL/min following the protocol described by Bian et al. ¹⁷ This began with 2% B for 0–5 min followed by a linear gradient from 2% to 4% B from 5 to 6 min; a linear gradient from 4% to 24% for 6–86 min then 24% to 44% B for 86–93 min and 44% to 90% B from 93 to 95 min; 90% B was used for 95–96 min followed by a gradient from 90% to 10% B for 96–98 min, isocratic at 10% B for 98–99 min at a flow rate of 450 nL/min, 10% to 90% B for 99–102 min at a flow rate of 600 nL/min; 90% B for 102–105 min, 90 to 2% B from 105 to 107 min, 2% B for 107–117 min, 2% B for 117–118 min at a flow rate of 450 nL/min, 2% B for 118–119 min at a flow rate of 300 nL/min, and finally 2% B from 119 to 120 min.

A stainless steel nano‐bore emitter (Thermo Fisher, ES542) provided the Thermo Orbitrap Fusion mass spectrometer with six gas‐phase fractions of the pooled samples using 4 Da fully staggered windows giving mass to charge ratios of 400–900 m/z in 100 m/z increments. Data were acquired in 16 Da windows from 400 to 1000 m/z in positive ion mode using data‐independent acquisition (DIA). The survey scans were done in centroid mode, with the maximum injection time set to auto and an automatic gain control target of 100,000 ions. The S‐lens RF level was set to 60. Isolation was performed in the quadrupole MS, and High‐energy Collisional Dissociation (HCD) MS/MS acquisition was performed in profile mode using a target of 500,000 ions.

The raw data were demultiplexed with 10 ppm accuracy after peak picking. The resulting peptides were searched and quantified via DIA‐NN (https://github.com/vdemichev/DiaNN) using a peptide length range of 7–50 amino acids, 2 missed cleavages, 3 variable modifications, clip N‐term M on, fixed C carbamidomethylation, variable modifications of methionine oxidation and n‐terminal acetylation, MS1 and MS2 accuracy set to 20 ppm, 1% FDR, and DIANN quantification strategy set to Robust LC (high accuracy). The resulting files were searched against a database of Chinese hamster proteins from Uniprot (December 2023).

Statistical analysis was performed using Fragpipe‐Analyst. The raw data was initially filtered to remove proteins that were identified as contaminants, reverse sequences, or only by site modifications to ensure the quality and relevance of the data.

2.4. Data interpretation

Proteins with high false discovery rate (FDR) confidence were linked to their master accession numbers in the UniProt protein database. The isoelectric point and GRAVY number for each protein were then evaluated using Expasy ProtParam (Expasy–ProtParam) based on the complete amino acid sequence. The abundance of individual host cell proteins was calculated based on the amount of protein used for lyophilization of that sample. Host cell protein transmission was calculated as the ratio of adjusted abundance of that protein in the permeate samples to the corresponding value in the feed sample.

3. RESULTS AND DISCUSSION

3.1. HCP clearance

Figure 1 shows the HCP breakthrough curve for a depth filtration experiment performed with the X0SP depth filter at a constant filtrate flux of 150 L/m²/h. The data are plotted as a function of the volumetric throughput, defined as the cumulative filtrate volume normalized by the depth filter area. The transmembrane pressure remained nearly constant throughout the filtration (at a value <11 kPa) due to the low turbidity of the HCP feed prepared by centrifugation followed by sterile filtration. The initial feed had an HCP concentration of 370 mg/L as determined by the Qubit protein assay, with the HCP transmission evaluated from the ratio of the total HCP concentration in the permeate samples to that in the feed. There appears to be a low level of HCP even in the first permeate sample (obtained at 40 L/m²), although this could also be due to interference in the assay associated with leachables from the depth filter media. ⁶ The total HCP concentration in the permeate was 50% of that in the feed at a throughput of approximately 220 L/m². This is similar to results reported in the literature for HCP removal by the X0SP depth filter ¹⁸ for experiments performed with a mAb‐producing CHO cell line. Note that data obtained in this work with HCP mixed with a purified monoclonal antibody (mAb), yielding a solution with 250 ppm HCP and 5 g/L mAb concentration, showed nearly complete mAb breakthrough after a volumetric throughput of only 25 L/m² with negligible HCP transmission out to at least 100 L/m² (see Appendix A).

FIGURE 1.

Breakthrough curves for total HCP (determined by Qubit assay) and three predominant proteins in the HCP feed: Thyrotropin‐releasing hormone receptor, inorganic pyrophosphatase, and phosphopyruvate hydratase (all determined by LC‐MS).

In addition to the results for total HCP, Figure 1 also shows data for 3 individual HCP identified by LC‐MS: phosphopyruvate hydratase, inorganic pyrophosphatase, and thyrotropin‐releasing hormone receptor. These 3 proteins were all in the top 350 by mass abundance of the proteins identified in the LC‐MS analysis; they were chosen for plotting in Figure 1 based on the large differences in binding behavior. The breakthrough curve for phosphopyruvate hydratase is similar to that for the total HCP, but the results for the inorganic pyrophosphatase and the thyrotropin‐releasing hormone receptor are considerably different. For example, the permeate sample obtained at 150 L/m² has nearly undetectable levels of thyrotropin‐releasing hormone receptor, while the inorganic pyrophosphatase is present at more than 40% of the concentration of that protein in the feed. Note that the breakthrough curves of these three HCPs do not correlate with their feed concentrations; the phosphopyruvate hydratase had the lowest abundance of the three proteins despite having intermediate breakthrough behavior.

3.2. LC‐MS analysis

The LC‐MS analysis identified more than 3000 individual HCP in the feed with a false discovery rate (FDR) threshold of less than 0.01. Although it is possible to construct breakthrough curves for each of these proteins, analogous to what was done in Figure 1, it is nearly impossible to interpret the results from a plot with that many breakthrough curves. Instead, the LC‐MS data were analyzed as follows. The fractional transmission was evaluated for each HCP in the permeate sample obtained through the X0SP filter at a volumetric throughput of 200 L/m², which is typical of the capacities of depth filters used for initial clarification (after centrifugation). The results for the 350 most abundant proteins are shown in Figure 2 as a scatter plot, with the data plotted as a function of the protein isoelectric point (pI) and GRAVY number. The color of the symbols shows the fractional transmission of each protein; high values (blue or black color) thus represent HCP with high transmission corresponding to low binding capacity (i.e., lower removal rates).

FIGURE 2.

Scatter plot of the 350 most abundant HCP through the X0SP filter at a volumetric throughput of 200 L/m². Color legend shows the transmission of each protein from 0 (white) to 1 (black).

All of the proteins with pI > 9.5 show less than 10% transmission, indicating that the very positively‐charged HCP are almost completely removed by the X0SP filter out to a volumetric throughput of 200 L/m². This behavior is consistent with the very high binding capacity for the model protein Ribonuclease A (pI = 9.6) observed previously for the X0SP filter. ⁶ In addition, the most hydrophilic proteins (the proteins with the most negative GRAVY numbers) also appear to show very low transmission. However, beyond these trends at very high pI and very negative GRAVY numbers, there appears to be very little correlation between the transmission and the protein pI or hydrophobicity. A multiple linear regression analysis for the transmission as a function of pI and GRAVY number gave an R² ≈ 0.31. Inclusion of the protein molecular weight only improved the fit to R² = 0.35. Instead, HCP binding to the X0SP filter appears to be very protein specific, likely due to the presence of specific “patches” on the protein surface that lead to strong interactions with the multiple elements (fibers, filter aid, and binder) that constitute the depth filter media. Note that the fractional transmission is also likely related to the actual concentration of the specific HCP in the feed, with those proteins present at high concentrations being more likely to saturate the available binding sites on the depth filter compared to HCP that are present at trace concentrations.

Previous studies of protein binding to the X0SP filter showed a significant contribution due to electrostatic interactions. ⁶ This was explored experimentally by resuspending the CHO cell pellet in 0.3× PBS instead of 1× PBS and using that low conductivity protein solution as the feed to the X0SP depth filter. To highlight the effects of the change in buffer conductivity, data are shown in Figure 3 only for individual proteins in which there was a significant difference in the fractional transmission between the 1× and 0.3× PBS buffers (in this case defined as C/C_Feed values that differ by at least 0.3). Only 19 of the 350 proteins showed this large a difference in fractional transmission. Of those proteins, 15 (79%) showed greater transmission (i.e., less binding) in the 1× PBS. The 4 proteins that had lower transmission (greater binding) in the 1× PBS all had pI between 4.9 and 6.1, that is, they are all negatively charged at the pH 7.4 used in these experiments. Note that the X0SP media has a net average zeta potential of −28 ± 2 mV ⁶ ; thus, electrostatic attraction would be most significant for proteins with moderate to high pI (or with positively‐charged patches on the protein surface). Interestingly, 2 of the 4 proteins that show greater binding at 1× PBS are relatively hydrophobic (GRAVY numbers greater than −0.2), suggesting that hydrophobic interactions may also play a role in protein binding to the X0SP filter.

FIGURE 3.

Scatter plot showing transmission behavior of individual HCP in 1× and 0.3× PBS through the X0SP filter at a volumetric throughput of 200 L/m². Yellow circles show those proteins for which the difference in C/C_Feed is at least 0.3 higher in 1× PBS; blue circles show those proteins with higher transmission in the 0.3× PBS buffer.

In addition to having very different physical properties (pI, MW, hydrophobicity, etc), several previous studies have identified what are typically referred to as “problematic” or “high‐risk” HCP. ¹⁹ , ²⁰ These individual proteins may be immunogenic, have specific biological activity that may be of concern to the patient, or may have enzymatic activity that can lead to degradation of the therapeutic protein. ²¹ , ²² The latter include both proteases as well as lipases, with some lipases causing significant degradation of polysorbate that is commonly used as an excipient in monoclonal antibody formulations. ²³ , ²⁴ Jones et al. ²⁵ have compiled a list of problematic/high‐risk HCP based on discussions within a BioPhorum industry group.

Table 1 summarizes the fractional transmission of twenty problematic HCP at a volumetric throughput of 100 L/m² for the X0SP and X0HC depth filters. The transmission values range from 0 to 1 for both filters, again highlighting the unique behavior of the individual HCP. Both depth filters provided almost complete removal of Lipoprotein Lipase, and the X0SP filter also provided almost complete removal of Lysosomal Acid Lipase (LAL); both of these lipases have previously been shown to degrade polysorbate and release free fatty acids, leading to particle formation. ²⁶ , ²⁷ Thus, the use of these depth filters could provide significant protection against polysorbate degradation in monoclonal antibody formulations. In contrast, neither filter provided any significant removal of Alpha‐enolase (2‐phospho‐D‐glycerate hydro‐lyase), Carboxypeptidase D, Glutathione S‐transferase, or Phospholipase B‐like 2.

TABLE 1.

Fractional transmission for problematic HCP through the X0SP and X0HC at a volumetric throughput of 100 L/m².

Protein ID	pI	GRAVY	MW	Transmission (X0HC)	Transmission (X0SP)
78 kDa glucose regulated protein (GRP78, BiP)	5.07	−0.481	72.4	0.28	0.01
Alpha‐enolase (2‐phospho‐D‐glycerate hydro‐lyase)	5	−0.319	15.5	1.00	0.00
Alpha‐enolase (2‐phospho‐D‐glycerate hydro‐lyase)	5	−0.319	15.5	1.00	1.00
Carboxypeptidase D (Cpd)	5.79	−0.407	123.4	1.00	1.00
Cathepsin B (CatB)	5.73	−0.348	37.5	0.44	1.00
Cathepsin L (CatL)	6.75	−0.463	37.3	0.26	0.00
Cathepsin Z (CatZ)	7.52	−0.452	34	0.87	0.01
Clusterin (CLU)	5.51	−0.637	51.8	0.17	0.00
Glutathione S‐transferase P 1 (GSTP1)	8.24	−0.272	25	1.00	1.00
Lipoprotein lipase (LPL)	7.95	−0.391	50.5	0.01	0.00
Lysosomal acid lipase (LAL)	7.32	−0.169	45.6	0.11	0.00
Matrix metalloproteinase‐19 (MMP‐19)	7.71	−0.382	58.9	0.76	1.00
Monocyte chemoattractant protein‐1 (MCP‐1)	9.32	−0.16	15.9	1.00	0.43
Phospholipase B‐like 2 (PLBL2)	5.9	−0.149	65.5	1.00	1.00
Procollagen‐lysine 2‐oxoglutarate 5‐deoxygenase_1 (PLOD1)	5.79	−0.409	76	0.06	0.00
Protein S100‐A6 (S10A6)	5.3	−0.253	10	1.00	0.39
Pyruvate kinase (PK)	7.58	−0.126	51.6	0.86	0.01
Serine protease (HTRA1)	6.54	−0.152	28.7	0.13	0.00

The X0SP showed equal or lower transmission (i.e., greater binding) than the X0HC filter for 18 out of 20 problematic HCP (90%). The two exceptions were Cathepsin B and Matrix metalloproteinase‐19 (MMP‐19), both of which showed 100% transmission through the X0SP. The lower transmission of the Matrix metalloproteinase‐19 through the X0HC filter may be related to the metal binding properties of this protein. Previous data with ovotransferrin, which has an iron binding site, showed unusually high binding capacity when using the X0HC depth filter media, ⁶ likely due to the presence of trace metals in the diatomaceous earth. ²⁸ Additional studies using purified Matrix metalloproteinase‐19 would be required to test this hypothesis.

The different binding characteristics of the X0SP and X0HC depth filters are examined in more detail in Figure 4 which shows a scatter plot highlighting those proteins that show differences in transmission greater than ±0.3 between the two filters at 100 L/m² volumetric throughput. The majority of these proteins (108 out of 114 proteins) had lower transmission (greater binding) with the X0SP filter, consistent with the results for the problematic HCP in Table 1. Similar results were obtained previously using a series of model proteins having a range of pI and GRAVY numbers, with all 7 model proteins showing greater binding capacities when using the X0SP filter. ⁶

FIGURE 4.

Scatter plot showing transmission behavior of individual HCP through the X0SP and X0HC depth filters at a volumetric throughput of 100 L/m². Yellow circles show those proteins for which the difference in C/C_Feed is at least 0.3 higher for the X0SP; blue dots show those proteins with higher transmission for the X0HC.

4. CONCLUSIONS

This study provides the first detailed analysis of HCP removal during depth filtration using LC‐MS, with data obtained with two model commercial depth filters (the X0SP and X0HC). The breakthrough curves for individual HCP were highly variable, with some HCP showing much earlier breakthrough and others much greater removal than the total HCP determined by a Qubit total protein assay. Scatter plots were constructed showing the HCP transmission as a function of the protein isoelectric point (pI) and hydrophobicity (GRAVY), with the data showing minimal correlation with either of these properties. However, data for the X0SP filter showed significantly greater binding of a number of HCP at low conductivity, demonstrating the importance of electrostatic interactions on HCP removal by the X0SP depth filter, particularly for HCP with higher isoelectric points (more positively‐charged).

The methodology presented in this work can also be used to compare the performance of different depth filters and to analyze the removal of problematic and challenging HCP, including HCP that may co‐elute with the product in the downstream chromatography steps. The X0SP was able to provide nearly complete removal of 12 of the 20 problematic HCP examined in this work (≤1% transmission at 100 L/m²), while the X0HC filter showed high removal of only 1 of the 20 problematic HCP. The use of LC‐MS allows for precise identification and quantification of individual HCP, highlighting the wide variability in breakthrough behavior observed during depth filtration. These findings underscore the complexity of HCP interactions with different depth filtration media and emphasize the need for greater understanding of the performance characteristics to improve the effectiveness of the depth filtration step in the downstream processing of biopharmaceutical products.

AUTHOR CONTRIBUTIONS

Liang‐Kai Chu: Data curation; formal analysis; investigation; writing – original draft. Ehsan Espah Borujeni: Conceptualization; funding acquisition; writing – review and editing. Xuankuo Xu: Conceptualization; funding acquisition; writing – review and editing. Andrew L. Zydney: Project administration; supervision; conceptualization; funding acquisition; writing – review and editing.

FUNDING INFORMATION

Financial support was provided by a grant from Bristol Myers Squibb.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

APPENDIX A.

In addition to the experiments with the HCP alone, a small number of experiments were performed in which the HCP (obtained from the null CHO cell line) was mixed with a highly purified monoclonal antibody (mAb). Typical data are shown in Figure A1 for a feed stream containing 250 ppm of HCP (relative to the mAb) and 5 g/L mAb concentration using the X0HC depth filter. The mAb concentrations were determined by Qubit assay (since the HCP levels were very small compared to the mAb) with the HCP concentration determined by a CHO HCP ELISA. The mAb shows nearly complete breakthrough after only 25 L/m² due to the very high mAb concentration in the feed (which leads to rapid saturation of the filter). The HCP concentration remained below 30 ppm (less than 12% transmission) over the 100 L/m² filtration; data could not be obtained at higher loadings due to the lack of feed material for these experiments. These results suggest that the presence of the mAb does not significantly enhance HCP breakthrough with these depth filters.

FIGURE A1.

Breakthrough curves obtained with a feed containing 250 ppm of HCP and 5 g/L of a mAb using the X0HC depth filter.

Chu L‐K, Borujeni EE, Xu X, Zydney AL. Use of LC‐MS to characterize host cell protein removal during depth filtration. Biotechnol. Prog. 2025;41(5):e70044. doi: 10.1002/btpr.70044

DATA AVAILABILITY STATEMENT

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