Host cell proteins (HCPs) are among the process-related impurities monitored during recombinant protein pharmaceutical process development. The challenges of HCP detection include (1) low levels of residual HCPs present in large excess of product protein, (2) the assay must measure a large number of different protein analytes, and (3) the population of HCP species may change during process development. Suitable methods for measuring process-related impurities are needed to support process development, process validation, and control system testing. A multi-analyte enzyme-linked immunosorbent assay (ELISA) is the workhorse method for HCP testing due to its high throughput, sensitivity and selectivity. However, as the anti-HCP antibodies, the critical reagents for HCP ELISA, do not comprehensively recognize all the HCP species, it is especially important to ensure that weak and non-immunoreactive HCPs are not overlooked by the ELISA. In some cases limited amount of antibodies to HCP species or antigen excess causes dilution-dependent non-linearity with multi-product HCP ELISA. In our experience, correct interpretation of assay data can lead to isolation and identification of co-purifying HCP with the product in some cases. Moreover, even if the antibodies for a particular HCP are present in the reagent, the corresponding HCP may not be readily detected in the ELISA due to antibody/antigen binding conditions and availability of HCP epitopes. This report reviews the use of the HCP ELISA, discusses its limitations, and demonstrates the importance of orthogonal methods, including mass spectrometry, to complement the platform HCP ELISA for support of process development. In addition, risk and impact assessment for low-level HCPs is also outlined, with consideration of clinical information.

The characteristics of protein detection and quantitation with SYPRO Ruby protein gel stain in one- and two-dimensional polyacrylamide gels were evaluated. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analyses of three different purified recombinant proteins showed that the limits of detection were comparable to the limits of detection with ammoniacal silver staining and were protein-specific, ranging from 0.5 to 5 ng. The linearity of the relationship between protein level and SYPRO Ruby staining intensity also depended on the individual protein, with observed linear dynamic ranges of 200-, 500-, and, 1000-fold for proteins analyzed by SDS-PAGE. SYPRO Ruby protein gel stain was also evaluated in two-dimensional electrophoretic (2-DE) analysis of Escherichia coli proteins. The experiment involved analysis of replicates of the same sample as well as dilution of the sample from 0.5 to 50 nug total protein across gels. In addition to validating the 2-DE system itself, the experiment was used to evaluate three different image analysis programs: Z3 (Compugen), Progenesis (Nonlinear Dynamics), and PDQuest (Bio-Rad). In each program, we analyzed the 2-DE images with respect to sensitivity and reproducibility of overall protein spot detection, as well as linearity of response for 20 representative proteins of different molecular weights and pI. Across all three programs, coefficients of variation (CV) in total number of spots detected among replicate gels ranged from 4 to 11%. For the 20 representative proteins, spot quantitation was also comparable with CVs for gel-to-gel reproducibility ranging from 3 to 33%. Using Progenesis and PDQuest, a 1000-fold linear dynamic range of SYPRO Ruby was demonstrated with a single known protein. These two programs were more suitable than Z3 for examining individual protein spot quantity across a series of gels and gave comparable results.

During production of a humanized antibody fragment secreted into the periplasm of Escherichia coli, proteolytic degradation of the light chain was observed. In order to determine which protease(s) were responsible for this degradation, we compared expression of the F(ab')(2) antibody fragment in several E. coli strains carrying mutations in genes encoding periplasmic proteases. Analysis of strains cultured in high cell density fermentations showed that the combination of mutations in degP prc spr was necessary for the cells to produce high levels of the desired recombinant antibody fragment. In order to eliminate the possible effects of mutations in other genes, we constructed E. coli strains with protease mutations in isogenic backgrounds and repeated the studies in high cell density fermentations. Extensive light chain proteolysis persisted in degP strains. However, light chain proteolysis was substantially decreased in prc and prc spr strains, and was further decreased with the introduction of a degP mutation in prc and prc spr mutant strains. These results show that the periplasmic protease Prc (Tsp) is primarily responsible for proteolytic degradation of the light chain during expression of a recombinant antibody fragment in E. coli, and that DegP (HtrA) makes a minor contribution to this degradation as well. The results also show that spr, a suppressor of growth defects in prc strains, is required for a prc mutant to survive throughout high cell density fermentations.

To understand the diversity in the cell culture harvest (i.e., feedstock) provided for downstream processing, we compared host cell protein (HCP) profiles using three Chinese Hamster Ovary (CHO) cell lines in null runs which did not generate any recombinant product. Despite differences in CHO lineage, upstream process, and culture performance, the cell lines yielded similar cell-specific productivities for immunogenic HCPs. To compare the dynamics of HCP production, we searched for correlations between the time-course profiles of HCP (as measured by multi-analyte ELISA) and those of two intracellular HCP species, phospholipase B-like 2 (PLBL2) and lactate dehydrogenase (LDH). Across the cell lines, proteins in the day 14 supernatants analyzed by two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) showed different spot patterns. However, subsequent analysis by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) indicated otherwise: the total number of peptides and proteins identified were comparable, and 80% of the top 1,000 proteins identified were common to all three lines. Finally, to assess the impact of culture viability on extracellular HCP profiles, we analyzed supernatants from a cell line whose viability dropped after day 10. The amounts of HCP and PLBL2 (quantified by their respective ELISAs) as well as the numbers and major populations of HCPs (identified by LC-MS/MS) were similar across days 10, 14, and 17, during which viabilities declined from ∼80% to <20% and extracellular LDH levels increased several-fold. Our findings indicate that the CHO-derived HCPs in the feedstock for downstream processing may not be as diverse across cell lines and upstream processes, or change as dramatically upon viability decline as originally expected. In addition, our findings show that high density CHO cultures (>10(7) cells/mL)-operated in fed-batch mode and exhibiting high viabilities (>70%) throughout the culture duration-can accumulate a considerable amount of immunogenic HCP (∼1-2 g/L) in the extracellular environment at the time of harvest (day 14). This work also demonstrates the potential of using LC-MS/MS to overcome the limitations associated with ELISA and 2D-PAGE for HCP analysis.

Two-dimensional electrophoretic analyses of Escherichia coli cells producing recombinant human growth hormone (Nutropin) in fermentations were conducted. The resulting two-dimensional protein profiles were compared with those of nonproducing (blank) cells. A qualitative comparison was performed to address regulatory issues in the biopharmaceutical industry, and a semiquantitative comparison was performed to reveal information about the physiological state of the cells. The protein spots unique to production fermentation profiles were all related to recombinant human growth hormone (hGH); these included intact hGH, charge variants of hGH, and a proteolytically cleaved form of hGH, as expected. There were no E. coli host cell proteins unique to either the production or blank fermentation profiles. Rather, all detectable differences in E. coli proteins were quantitative in nature. Specifically, the levels of IbpA (inclusion body binding protein A), Ivy (inhibitor of vertebrate lysozyme), and a cleaved form of GroEL (Hsp60 homolog) were higher in hGH production profiles, whereas the levels of GlmU protein and PspA (phage shock protein A) were higher in blank profiles. In general, the high degree of similarity between proteomes for hGH-producing and nonproducing cells suggests that E. coli proteins from a nonproducing (blank) fermentation are appropriate for eliciting antibodies that are then used in immunoassays to measure host cell proteins in samples from production fermentations.