Identification and Monitoring of Host Cell Proteins by Mass Spectrometry Combined with High Performance Immunochemistry Testing

Biotherapeutics are often produced in non-human host cells like Escherichia coli, yeast, and various mammalian cell lines. A major focus of any therapeutic protein purification process is to reduce host cell proteins to an acceptable low level. In this study, various E. coli host cell proteins were identified at different purifications steps by HPLC fractionation, SDS-PAGE analysis, and tryptic peptide mapping combined with online liquid chromatography mass spectrometry (LC-MS). However, no host cell proteins could be verified by direct LC-MS analysis of final drug substance material. In contrast, the application of affinity enrichment chromatography prior to comprehensive LC-MS was adequate to identify several low abundant host cell proteins at the final drug substance level. Bacterial alkaline phosphatase (BAP) was identified as being the most abundant host cell protein at several purification steps. Thus, we firstly established two different assays for enzymatic and immunological BAP monitoring using the cobas® technology. By using this strategy we were able to demonstrate an almost complete removal of BAP enzymatic activity by the established therapeutic protein purification process. In summary, the impact of fermentation, purification, and formulation conditions on host cell protein removal and biological activity can be conducted by monitoring process-specific host cell proteins in a GMP-compatible and high-throughput (> 1000 samples/day) manner.


Introduction
Host cell proteins (HCPs) carry potential clinical safety risks for patients treated with biologics. On the one hand, HCP might cause an immune response (due to their "non-self" nature), adjuvant activity, and theoretically also function in the human body [1][2][3]. Furthermore, HCPs with protease activity have the potential to impact product stability [4]. Consequently, regulatory guidelines mandate the setting of HCP specifications [5].
Thus, one key aspect of any biologics manufacturing is to reduce HCP to levels considered acceptable in the final drug [6]. The HCP composition is impacted by the proteome complexity of the utilized host expression system [7][8][9], the manner in which the therapeutic protein is expressed [10][11][12][13], and the purification process itself [10,12]. Moreover, all methods for analytical HCP characterization face challenges due to the dynamic range of HCP abundance at proteome and final drug level.
Several analytical techniques have been used for the detection, identification, and quantification of HCPs [1,3,14,15]. To perform bio-process and release analytics, immunoassays like protein gel blots and multicomponent generic or processspecific enzyme-linked immunosorbent assays (ELISA) are most commonly used to detect and monitor HCPs [16][17][18]. The ELISA technique is typically applied for HCP analysis, mostly due to the good precision of the method and also that it provides quantitative results for setting control limits and specifications. However, generic ELISAs do not offer complete coverage for all process-specific HCPs and process-specific ELISAs might be not qualified to evaluate the HCP content after process changes [3,[16][17][18]. Two-dimensional gel electrophoresis combined with fluorescent staining is also applied for the detection and quantification of HCPs [19,20]. The technique is semi-quantitative, has a limited dynamic range, and needs mass spectrometry for HCP identification.
In the present study, an approach employing affinity chromatography to capture HCPs, highly sensitive LC-MSMS, and high throughput immunoassay testing for the enrichment, identification and quantification of HCPs in biotherapeutics was developed. This test system allowed us to identify and monitor Bacterial Alkaline Phosphatase in a biopharmaceutical purification process.

Results
Increased levels of HCPs were detected in the manufacturing process for a recombinant protein derived from E. coli by ELISA and RP-HPLC analysis. At final drug substance level, several batches with HCP levels minimal greater than the release specification of 30 ppm were observed by a process-specific ELISA system. RP-HPLC analysis with UV detection is routinely applied to monitor product variants at the final drug substance level and at various purifications steps (Figure 1). At the final drug substance level no significant differences in product purity where observed for batches with elevated HCP levels ( Figure 2A). The first chromatographic purification step of the recombinant protein is accomplished by metal chelate chromatography (purification step 1). At this stage several product variants (marked by asterisks) can be observed in the obtained elution pool ( Figure 2B).
In general, no significant differences in product variants where observed for batches with elevated HCP levels at purification step 1 level. However, a slight but distinct increase in peak intensity of the product variant with a retention time of 25 min could be observed for batches with elevated HCP levels ( Figure 2B; marked by an arrow). The existence of elevated product variants or HCP levels was further suggested by SDS-PAGE analysis of the respective HPLC fractions (retention time window: 24-26 min), in which an increased content of a protein species with a molecular weight of around 23 kDa was observed for batches with elevated HCP levels ( Figure 3). The unknown protein was identified as E. coli Alkyl hydroperoxide reductase subunit C (see Table 1) by nano-ESI-MS peptide mapping after in tryptic-gel digestion combined with database searching (mass spectrometric data not shown). Additionally, various potential host cell proteins, not detected by RP-HPLC analysis, were visualized by SDS-PAGE analysis of total purification step 1 elution pool ( Figure 4). The dominant HCP with a molecular weight of around 50 kDa was identified as Bacterial Alkaline Phosphatase (BAP; see Table 1) by the procedure as described above (data not shown) and located in the injection peak of the RP-HPLC chromatogram by HPLC fractionation (retention time window: 2-4 min, Figure 2B). The SDS-PAGE analysis of purification step 1-3 elution pools suggest a complete HCP removal by the applied purification process ( Figure 4). Nevertheless, tryptic peptide mapping combined with comprehensive online liquid chromatography mass spectrometry (LC-MS/MS) was utilized to identify low abundant HCPs of the different elution pools. Protein identification by database searching was successful to identify various bacterial proteins present in the elution pools of the purifications steps 1 and 2. The identified HCPs are summarized in Table 1. Although the analysis was performed with a highly sensitive LTQ Orbitrap Velos electrospray mass spectrometer, only one bacterial protein (BAP) was verified in the elution pool of purification step 3 and no HCPs were detected at the final drug substance level (Table 1). Since the deployed process-specific ELISA system does demonstrate low abundant HCP levels at final bulk stage ( Figure 2), we introduced HCP enrichment by affinity chromatography using the ELISAs polyclonal antibodies in our sample preparation protocol to detect bacterial proteins of low abundance (see details in materials and methods). In order to judge if the HCP enrichment step does significantly enrich or deplete individual HCPs, purification step 1 elution pool was again analyzed with and without applying the affinity enrichment procedure. The identified proteins (identification score ≥ 20; peptides with an FDR <1%) are listed in Table 2 (Repeatability results are summarized in Table S1). All abundant bacterial proteins of the purification step 1 elution pool were also detected after applying the HCP enrichment step. Although the identification scores of some bacterial proteins do suggest alterations of the relative protein abundances only minor effects on protein sequence coverage were observed. In addition, an increasing number of identified HCPs was demonstrated as a consequence of the affinity enrichment procedure. Thus, from a qualitative point of view, the results achieved do not suggest significant alterations of the HCP profile related to the affinity enrichment step.
Next, we employed the described approach to identify HCPs at the final drug substance level. A total of 12 bacterial proteins (with an identification score ≥ 20), not detected without affinity enrichment, were identified at final bulk level (Table 3). Eight proteins were already verified at purification step 1 elution pool level (including BAP and Alkyl hydroperoxide reductase subunit C). However, the Protein tolB, DNA protection during starvation protein, 2-amino-4-hydroxy-6-hydroxymethyldihydropteridine pyrophosphokinase, and N-acetylmuramoyl-L-alanine amidase AmiD were only detected at final drug substance level suggesting a relative enrichment of the protein through the purification process.
In summary, the application of affinity chromatography combined with comprehensive LC-MS was adequate to identify low abundant HCPs at the final drug substance level.
Since BAP was identified as being the most abundant HCP at purification step 1 elution pool level and is still traceable at final drug substance level we developed two different assays for enzymatic and immunological BAP monitoring on a cobas INTEGRA® 400 plus and a cobas e 411 system, respectively 1 . The method validation results for both cobas ® systems demonstrated acceptable analytical performance and are in accordance with the manufacturer's accuracy and imprecision criteria ( Table 4). The cobas® analysis of all total purification step elution pools and the final drug substance material does demonstrate an almost complete BAP removal by the applied purification process (Table 5). At final drug substance level the detected immunological BAP activity of 0.1 U/mg was close to the quantification limit (0.12 U/mL) and the observed enzymatic BAP activity was below the detection limit of 0.2 mU/mL (see Table 4). As initially described, several batches with increased HCP levels were observed at final drug substance level by a process-specific ELISA system. To verify if BAP does contribute to the elevated total HCP levels the enzymatic and immunological BAP activity of various batches was analyzed at purification step 3 elution pool level. The results are  Table 6 and demonstrate that no causal relationship between elevated total HCP levels and enzymatic and immunological BAP activity is verifiable.

Discussion
For the identification and quantification of HCPs in biotherapeutics, a test system employing HCP affinity enrichment, highly sensitive LC-MS, and high throughput immunoassay testing was developed. Direct LC-MS analysis was capable of identifying 76 HCPs at purification step 1 level but did not verify the presence of HCPs at the final drug substance level. In contrast, the application of affinity chromatography for HCP enrichment combined with LC-MS was adequate to identify 85 HCPs at purification step 1 level. In addition, the results achieved do not suggest, from a qualitative point of view, significant alterations of the HCP profile related to the affinity enrichment step. The described approach resulted in the confirmation of 12 low abundant HCPs in the final drug substance. All identified bacterial HCPs have the potential to trigger an immune response in human, although the respond will depend on composition and amount of bacterial proteins administered [2,3]. Moreover, five low abundant bacterial proteins with possible catalytic activity were verified.
Bacterial Alkaline Phosphatase (BAP) is a widely distributed non-specific phosphomonoesterase that also catalyzes phosphoryl transfer reaction to various alcohols [31,32]. E. coli Alkyl hydroperoxide reductase (AhpC) directly reduces organic hydroperoxides in its reduced dithiol form and might act as an antioxidant enzyme [33]. 2-amino-4-hydroxy-6hydroxymethyldihydropteridine pyrophosphokinase, Adenylate kinase, and N-acetylmuramoyl-L-alanine amidase AmiD were also detected at the final drug substance level. The identified bacterial enzymes are involved in folic acid biosynthesis,  nucleotide biosynthesis, and cell wall biogenesis/degradation [34][35][36][37]. The described catalytic activities do not suggest a negative impact on product stability. However, bacterial enzymes could theoretically also have enzymatic effects in humans, although, to the author's knowledge, reports to this risk have not been published [3]. BAP, AhpC, and Adenylate kinase were detected at purification step 1 and final drug substance level whereas 2-amino-4-hydroxy-6hydroxymethyldihydropteridine pyrophosphokinase and Nacetylmuramoyl-L-alanine amidase AmiD were only detected at the final drug substance level. Thus, the data indicate a less effective removal or relative enrichment of both enzymes through the purification process.
Subsequently, we focused on developing test systems for high throughput monitoring of specific bacterial HCPs. Since BAP was identified as being the most abundant HCP at purification step 1 level, we firstly established two different assays for enzymatic and immunological BAP monitoring using the cobas® technology. The obtained results demonstrate an almost complete removal of BAP enzymatic and immunological activity by the applied purification process. However, since BAP enzymatic activity was verified until purification step 3 level BAP activity at the final drug substance level is most likely, although to a much lower extent. These results are in agreement with the data obtained by comprehensive LC-MS analysis, in which BAP at final bulk level was only found after enrichment by affinity chromatography. In addition, we assessed if BAP does contribute to the elevated total HCP levels of some drug substance batches. The results revealed that no causal relationship between elevated total HCP levels and enzymatic and immunological BAP activity is traceable. Currently, we are developing novel immunological and enzymatic cobas® test systems to assess AhpC, 2-amino-4hydroxy-6-hydroxymethyldihydropteridine pyrophosphokinase, Adenylate kinase, and N-acetylmuramoyl-L-alanine amidase AmiD removal by the applied purification process and perhaps more importantly, their catalytic activity at the final drug substance level. Recently, the identification and quantification of HCPs in biotherapeutics by 2D-LC-MS was demonstrated [29,30]. The described approaches allow the simultaneous quantification of various low abundance HCPs. In the present study, the application of affinity chromatography combined with comprehensive LC-MS analysis and cobas® technology was adequate to identify low abundant HCPs at final drug substance level and to subsequently monitor the enzymatic and immunological activity of BAP at various purifications steps. Accordingly, the impact of fermentation, purification, and formulation conditions on HCP removal and biological activity can be conducted by monitoring process-specific HCPs in a GMP-compatible and high-throughput (> 1000 samples/day) manner.

Reversed Phase Chromatography (RP-HPLC)
Reversed phase chromatography was performed on a Dionex Summit ® HPLC system (Thermo Fisher Scientific, Bremen, Germany) with UV detection at 210 nm. The separation was carried out on a YMC-Pack ODS-AQ analytical column (3 x 150 mm, S-3μ, 200 Å, carbon content: 11.0-11.5 %, YMC, Tokyo, Japan) between 20°C and 25°C. A step gradient using 0.1% TFA, 30 % acetonitrile as solvent A and 0.1 % TFA, 80 % acetonitrile as solvent B at 0.4 mL/min was applied. For the chromatographic analysis 6 µg of total protein was injected. Fractions were collected manually.

Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE)
One-dimensional SDS-PAGE was performed with Novex® 18% Tris-Glycine gels in a XCell SureLock® Mini-Cell (Life Technologies Corporation, Darmstadt, Germany). Samples were reduced with NuPAGE® Sample Reducing Agent (catalogue number: NP0004) and gel electrophoresis was

HCP identification by Nano ESI-MS/MS
For the Nano ESI-MS/MS analysis of the in-gel digests, peptides were desalted and concentrated with ZipTip® C18 Pipette Tips (Millipore Corporation, Billerica, USA) and eluted in 20 µL 1% formic acid, 50% acetonitrile. Nano ESI-MS/MS analyses were performed on a QTOF Ultima mass spectrometer (Waters, Manchester, U.K.) equipped with a TriVersa NanoMate (Advion, Inc., Ithaca, USA). Spectra were recorded in the positive ion mode. Sequencing was performed

In-solution digestion
For the proteolytic digestion, samples were denatured in 0.4 M Tris-HCl, 8 M guanidine hydrochloride, pH 8.5 by diluting 150 µg of protein in a total volume of 300 µL. For reduction, 10 µl of 0.1 g/mL dithiothreitol was added followed by incubation at 50°C for 1 hour. After alkylation of free cysteine by adding 10 µl of 0.33 g/mL iodoacetic acid and incubation at room temperature under exclusion of light for 30 min, the buffer was exchanged to digestion buffer (0.1 M Tris-HCl, pH 8.5) by application onto an illustra™ NAP™-5 gel filtration column (G-25, GE Healthcare, Buckinghamshire, UK). The NAP™-5 eluate (500 µL) was mixed with 30 µL of a solution of 0.1 mg/mL sequencing grade trypsin (Promega, Madison, USA) in 10 mM HCl and incubated at 37°C for 18 h.

HCP identification by LC-MS/MS
LC-MS/MS analysis of in solution digests were performed on a LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) coupled to an Acquity UPLC system (Waters, Manchester, UK). Peptides were separated on a BEH C18 column (1.7 µm 2.1x150 mm, Waters, Manchester,   -2000) were acquired by the Orbitrap system with a resolution of r = 30000. The ten most abundant peptide ions with charge states > +1 were sequentially isolated and fragmented with CID and a normalized collision energy of 40 V using helium as collision gas. The resulting fragment ions were detected by the ion trap. The automatic gain control (AGC) target values were set to 1*10 6 for Full MS scans in the Orbitrap mass analyzer and 1*10 4 for MS/MS scans in the ion trap mass analyzer.

LC-MS/MS database query
CID spectra were searched against an in-house E. coli database using Proteome Discoverer V1.3 (Thermo Fisher Scientific, Bremen, Germany). The custom protein database was compiled from 5,967 E.coli protein sequences (from actual Swiss Prot K12 database), the sequence of the therapeutic protein, and the sequence of porcine trypsin. The following SEQUEST® search parameters were applied: (1) trypsin with a maximum of one missed cleavage, (2) carboxymethylation of cysteines as fixed and (3) oxidation of methionine as variable modifications. Precursor mass tolerance was set to 10 ppm and the fragment mass tolerance was set to 0.8 Da. A False Discovery Rate (FDR) of 1% was calculated. Only proteins specified with a score value ≥ 20 and with high confidence unique peptides (FDR < 1%) were considered as positive hits. The mass spectrometry proteomics data for HCP identification at drug substance level have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository [38,39] with the dataset identifier PXD000509 and DOI 10.6019/PXD000509.

Quantitative determination of the enzymatic activity of alkaline phosphatase
The colorimetric assay was carried out on a cobas INTEGRA® 400 plus system using the ALP IFCC Gen.2 cassettes (Roche Diagnostics Ltd., Basel, Switzerland) according to the manufacturer's guidelines. Briefly, the principle of the assay is the cleavage of p-nitrophenyl phosphate by alkaline phosphatase into phosphate and p-nitrophenol. The pnitrophenol released is directly proportional to the catalytic alkaline phosphatase activity. The activity is determined by measuring the increase in absorbance at 409 nm. The assay is performed at 37°C. A minimal sample volume of 250 µL was required. Calibration was done with Bacterial Alkaline Phosphatase (Life Technologies GmbH, Darmstadt, Germany, Catalogue number 18011-015).

Bacterial Alkaline Phosphatase (BAP) ECLIA
The immunological BAP content was determined by an electrochemiluminescence immunoassay (ECLIA) on a cobas e 411 analyzer (Roche Diagnostics Ltd., Basel, Switzerland) according to the manufacturer's guidelines. Briefly, the assay is based on a sandwich principle using a mouse monoclonal anti-BAP antibody (in-house development). First, 30 µl of sample, biotinylated anti-BAP capture antibody and ruthenium-labeled anti-BAP detection antibody were incubated for 9 min forming a ternary complex. In a second step after addition of streptavidincoated microparticles, the complex is bound to the solid phase via interaction of biotin and streptavidin during 9 min of incubation. The reaction mixture is aspirated into the measuring cell where the microparticles are magnetically captured to the surface of the electrode. Voltage-induced chemiluminescence is measured by a photomultiplier. The concentration of BAP in the test sample is finally calculated from a BAP (Life Technologies GmbH, Darmstadt, Germany, Catalogue number 18011-015) standard curve of known concentration. Table S1. Repeatability of HCP identification results. Database query was conducted by analyzing LC-MS/MS CID spectra using Proteome Discoverer V1.3 and a false discovery rate FDR < 1% of three purification step 1 elution pool samples (three independent sample preparations). The Top 15 HCPs were sorted according to the score value of Analysis 1. Accession number, http://www.uniprot.org/; MW, theoretical molecular weight. (XLSX)