Ceramic hydroxyapatite chromatography plays a critical role in bispecific antibody purification process for impurity removal

Abstract

Background

Significant challenges exist in downstream purification of bispecific antibodies (BsAbs) due to the complexity of BsAb architecture. A unique panel of mispaired species can result in a higher level of product-related impurities. In addition to process-related impurities such as host cell proteins (HCPs) and residual DNA (resDNA), these product-related impurities must be separated from the targeted BsAb product to achieve high purity. Therefore, development of an efficient and robust chromatography purification process is essential to ensure the safety, quality, purity and efficacy of BsAb products that consequently meet regulatory requirements for clinical trials and commercialization.

Methods

We have developed a robust downstream BsAb process consisting of a mixed-mode ceramic hydroxyapatite (CHT) chromatography step, which offers unique separation capabilities tailored to BsAbs, and assessed impurity clearance.

Results

We demonstrate that the CHT chromatography column provides additional clearance of low molecular weight (LMW) and high molecular weight (HMW) species that cannot be separated by other chromatography columns such as ion exchange for a particular BsAb, resulting in ≥98% CE-SDS (non-reduced) purity. Moreover, through Polysorbate-80 (PS-80) spiking and LC–MS HCP assessments, we reveal complete clearance of potential PS-80-degrading HCP populations in the CHT eluate product pool.

Conclusions

In summary, these results demonstrate that CHT mixed-mode chromatography plays an important role in separation of product- and process-related impurities in the BsAb downstream process.

Statement of Significance: CHT mixed-mode chromatography demonstrates consistent, scalable performance and achieves optimal product variant and process-related impurity separation and resulting high-quality BsAbs.

INTRODUCTION

BsAbs are emerging as next-generation biotherapeutics attributed to their dual-targeting capability, specificity and reduced toxicity effects [1]. However, significant challenges exist in downstream purification of BsAbs due to the intricacy of BsAb structure [2]. Despite various engineering approaches to enable correct pairing, a unique panel of product-related variants can result in elevated levels of HMW aggregates as well as LMW species in contrast to traditional monoclonal antibodies (mAbs) (Fig. 1, Table 1) [3]. These comprise free heavy chains (HCs), free light chains (LCs), homodimers, half molecules and other mispaired entities (Fig. 1, Table 1) that can associate with product pools and share similar properties with the main heterodimer product, rendering them challenging to separate [3–5]. In addition to these product-related variants, process-related impurities such as HCPs and resDNA must be cleared through the downstream process to generate high-quality product. Moreover, potential species within the population of HCPs, including lipases, that can degrade PS-80 [6–8], a common excipient in final product formulation buffers, must be monitored and an effective downstream process applied to remove these species. Thus, BsAbs warrant a robust and scalable downstream process to ensure clearance of additional mispaired species and therefore the safety, quality and potency of BsAb products.

Figure 1.

BsAb structure. 1 + 1 IgG1 format targeting two distinct antigens, denoted Antigen1 and Antigen2. Pairing mutations are incorporated to enhance heterodimer formation. The heavy chain (HC) is comprised of the VH, CH1, hinge, CH2 and CH3 domains, and the light chain (LC) is composed of the VL and CL domains.

Table 1.

BsAb impurity profile. In addition to process-related impurities such as HCP, potential LMW and HMW mispaired entities can co-elute with BsAb product pools and must be monitored and cleared during the downstream process. LC = Light Chain; LL = Light Light; HC = Heavy Chain; HL = Heavy Light; HH = Heavy Heavy; HHL = Heavy Heavy Light; HHLL = Heavy Heavy Light Light

Different types of mixed-mode chromatography resins offer unique separation capabilities tailored to BsAbs and thus can serve as an optimal polishing step to enhance purification of BsAbs [4]. These include resins that employ both IEX and hydrophobic properties to promote differential separation [4]. In particular, CHT chromatography exhibits multimodal characteristics and has been utilized for aggregate and other impurity separation from mAbs and BsAbs [9–11]. CHT is a macroporous form of calcium phosphate that exploits a combination of cation exchange and metal affinity interactions to achieve separation of impurities from BsAb heterodimer [10, 12].

We incorporate CHT chromatography as a polishing step within a BsAb downstream purification process and demonstrate that it provides additional clearance of LMW and HMW species that cannot be separated by other chromatography columns. Moreover, we monitor and reveal complete clearance of potential PS-80-degrading HCP populations in the CHT eluate product pool. Incorporation of CHT mixed-mode chromatography in the BsAb downstream process achieves optimal purity for a 1 + 1 IgG1 BsAb and thus has the potential to be applied across various multi-specific antibody configurations and biologic modalities.

MATERIALS AND METHODS

CHT chromatography

CHT chromatography development was conducted using pre-packed 0.77 cm Inner Diameter (ID), 5 or 10 cm bed height (BH) CHT Type I columns (Bio-Rad Laboratories, Inc.) on an AKTA Avant purification system (Cytiva). Protein A-purified BsAb product pool was pH- and conductivity-adjusted prior to loading onto a CHT column for initial gradient experiments all under pH 7.5 conditions, and operated in bind and elute mode at a process residence time (RT) of 4 min at ambient temperature. Injection pulse methodology to execute phosphate and salt gradients were utilized to determine operational conditions for the wash and elution steps. From these experiments, a sodium phosphate concentration of 20 mM, with varying sodium chloride concentrations, was established as optimal for the wash and elution conditions.

For bench-scale and 250 L pilot-scale confirmation runs, CHT type I resin (Bio-Rad Laboratories, Inc.) was packed in 5 cm ID HiScale and BPG 20 cm ID column hardware (Cytiva), respectively, targeting a 20 cm BH. Pilot-scale runs were executed on an AKTA Ready system (Cytiva). All runs were operated at 4 min RT at ambient temperature, and a loading capacity of 20 g/L_resin was targeted.

Purity assessment

The purity profiles of pooled CHT eluate fractions were evaluated by a panel of assays including analytical size-exclusion chromatography-HPLC (SEC-HPLC), reverse-phase UPLC (RP-UPLC) and non-reduced (NR) capillary electrophoresis-SDS (CE-SDS) for characterization of product-related impurities.

Samples for SEC analysis were separated on a Tosoh TSKgel G3000SWXL 7.8 mm × 30 cm, 5 μm column, part number 08541, in 200 mM potassium phosphate, 250 mM potassium chloride, pH 6.2 mobile phase buffer at a flow rate of 0.5 mL/min and processed on a Waters HPLC system equipped with a UV–Vis detector set to a wavelength of 280 nm.

For RP-UPLC analysis, samples were diluted to 1.0 mg/mL in water, separated on a Bioresolve RP mAb polyphenyl 2.7 × 2.1 × 100 column (Waters) under 0.1% trifluoroacetic acid (TFA) mobile phase A and 0.075% TFA/acetonitrile mobile phase B gradient conditions, and processed on an UPLC/LWS system equipped with a photodiode array (PDA) detector.

Samples for CE-SDS were prepared according to IgG purity/heterogeneity assay kit instructions (AB Sciex Part Number A10663), run under non-reducing (NR) conditions in the presence of iodoacetamide (IAM), and processed on a PA 800 Plus Pharmaceutical Analysis System equipped with a PDA detector. Analyses for all the above assays were conducted using Waters Empower software.

Impurity analysis

Impurity profile assessments for HCP and residual Protein A (resProA) were performed using ELISA kits (Cygnus Technologies, Part Number F550 and RepliGen BioProcessing Part Number 9333-1, respectively), according to the manufacturer’s instructions. Samples for resDNA analyses were prepared using PrepSEQ Nucleic Acid Extraction and Residual DNA Sample Prep Kits (Applied Biosystems, Box 1 (4400793); Box 2 (4400787), Box 3 (4400675) and Prep kit (4399042), according to the manufacturer’s instructions, and resDNA quantified by qPCR using resDNASEQ CHO Real-Time PCR Reagents (Applied Biosystems Part Number 4402431).

PS-80 degradation assay

For PS-80 degradation assessments, all in-process samples were buffer-exchanged into 20 mM sodium acetate, pH 5.0 and concentrated to 40 g/L using an amicon ultra 15 (30 kDa) centrifugal filter, targeting at least 3 mL final volume to perform the subsequent sample preparation. Enzymatic activity is pH-dependent, so it is important to only compare samples that are in the same buffer matrix and pH [13]. In a biosafety cabinet, samples were sterile-filtered and spiked with a sterile stock of PS-80 and diethylenetriaminepentaacetic acid (DTPA) to a target concentration of 0.05% PS-80 and 5 μM DTPA. DTPA is a chelating agent, which is added to prevent PS-80 degradation due to oxidation. PS-80 can also be degraded by oxidation induced by light stress. Samples were well-mixed, aliquotted and incubated light protected at 37 °C to speed up enzymatic activity. An aliquot was frozen immediately and used as a control. Additional aliquots were frozen at multiple timepoints up to 7 days, then thawed together to analyze PS-80 content using UPLC with a charged aerosol detector (CAD). Polysorbate separation from the other matrix components was performed using a Waters Oasis Max, 2.1 × 20 mm, 30 cm column on Waters H Class UPLC system and quantified with a Corona Ultra RS CAD from Dionex (Thermo Fisher Scientific, Waltham, MA). Mobile phase compositions were 0.05% formic acid in water (A) and 100% IPA (B). Running conditions were 90% A for 1 minute, 80% A for 9.5 minutes, 100% B for 4.5 minutes (eluting the Polysorbate from the column), followed by 90% A for 5 minutes. Quantitative analysis of the chromatograms was performed with Waters Empower 3 software from a PS-80 standard curve. Samples were injected to a nominal concentration of 50 μg/mL. A bulk standard was prepared to a nominal concentration of 50 μg/mL and frozen; aliquots of this standard were injected during each experiment for system suitability testing and to track the long-term precision of the method. The UPLC and detector were set to the following parameters for experiments: column temperature 35 °C, autosampler temperature 5 °C, pump flow rate 1 mL/min, sample injection volume 25 μL, CAD gas pressure (nitrogen) 35 psi, CAD nebulizer temperature 25 °C, CAD range 200 pA. This approach is similar to others previously published, which used different modes of detection [14–16].

LC-mass spectrometry

Semi-quantitative analysis of high abundance HCP species was performed by LC-mass spectrometry (LC–MS). Native digestion and Hi-3 quantitation were used for detection, and samples were analyzed by an Orbitrap Lumos mass spectrometer. Sample preparation and LC–MS experimental methods were described previously [17], except MS parameters were adjusted to match the Lumos instrument [18]. Data were processed using ProgenesisQI for Proteomics with Byonic search engine. The protein database used for HCP searches included all CHO proteins in the Uniprot database (retrieved 1 October 2020), the sequence of the therapeutic, and an integrated “Common contaminant” library in the Byonic search engine (includes trypsin, keratins, etc.). Quantities were calculated by Progenesis relative to a yeast enolase internal standard (5000 fmol per 1 mg protein) and using sequence calculated masses for parts-per-million (ppm) conversion. Conflicting peptides were manually screened and removed, as well as low-confidence outliers. A separate search was performed in Byonic alone on each sample raw file for comparison of HCP hits and confidence scores.

RESULTS

The complexity of BsAb architecture results in higher levels of mispaired species during cell culture expression. Although engineering strategies with mutations to enable heterodimer formation are incorporated, several product-related impurities potentially exist in the harvest/clarified cell culture fluid prior to purification (Fig. 1, Table 1). Affinity capture chromatography alone as well as a single polishing step following affinity chromatography are not sufficient to achieve high bispecific heterodimer purity, more specifically SEC monomer and CE-SDS (NR) main peak purity of no less than (NLT) 90.0% with concomitant low SEC HMW of no more than (NMT) 5.0% and process-related impurity levels, or NMT 6.00 pg/mg resDNA and NMT 100 ng/mg HCP, aligned with CHO-derived biologic therapeutics purity specifications [19]. Robust polishing chromatography steps are essential for clearance of the multitude of these potential mispaired entities that can co-elute with BsAb product pools. Mixed-mode CHT chromatography as a polishing step was developed to remove these impurities in the affinity capture column elution pool. To maximize yield as well as impurity separation during the CHT column step, step gradient experiments were executed. A sodium chloride gradient in 50 mM increments was run under a constant sodium phosphate concentration (Fig. 2). The profile was characterized by an initial elution peak (Fraction1) that was exclusively half molecule impurity identified by RP-UPLC. The main elution peak (Fraction3) demonstrated high SEC-HPLC (98.7%), RP-UPLC (99.2%) and CE-SDS (NR) (96.7%) purity, with very low HMW aggregate levels <0.5% (Table 2). By CE-SDS (NR) analysis, LMW levels of 3.3% were also lowest in the main peak Fraction3 and more enriched in leading and tailing fractions (Table 2). Furthermore, HCP levels were elevated in the leading peak and tail regions (data not illustrated). Based on this profile, subsequent step gradient experiments confirmed the wash and elution conditions, containing a constant sodium phosphate concentration of 20 mM and sodium chloride concentrations within a range from 50 to 500 mM, depending on process conditions and molecule properties, achieved optimal HCP removal from product-related variants and generated high-purity product.

Figure 2.

CHT step gradient evaluation. AKTA chromatogram of step sodium chloride gradient elution reveals impurity separation in the leading initial peak and elution tail region.

Table 2.

CHT step gradient purity analysis. SEC-HPLC, RP-UPLC and CE-SDS (NR) analyses indicate high halfmer levels in the initial leading peak (1) as well as elevated impurity levels in the elution tail region (5). The main region of the elution peak (3, highlighted) exhibits highest purity. Furthermore, HCP levels are elevated in the leading peak and elution tail regions (not presented).

Next, process performance was confirmed at bench-scale. CHT step yield was 89% and overall process yield, not including harvest and UF/DF, was 63%. With implementation of CHT as the polishing step, satisfactory clearance of HCP process-related impurity, HMW variants and LMW variants was achieved (Table 3). Notably, there was further removal of LMW species proceeding from CHT load (4.9%) to CHT eluate (2.0%) by CE-SDS (NR) analysis (Table 3). Overall purity by SEC was 99.8% and by CE-SDS (NR) was 98.1% following CHT (Table 3).

Table 3.

Lab-scale and pilot campaign product quality summary. In a lab-scale carrythrough run, high SEC and CE-SDS (NR) purity were achieved following CHT chromatography, utilizing conditions optimized from gradient experiments. Purity assessment from a pilot scale-up run indicates high SEC and CE-SDS (NR) BsAb purity was achieved in the final CHT chromatography step. In addition, resDNA, HCP and resProA process-related impurity levels were well below the acceptable limits following CHT.

Assay		Parameters	Lab-scale		Pilot-scale
			CHT load	CHT eluate	CHT load	CHT eluate
Concentration	A280	Step yield %		89		82
Purity	SEC-HPLC	Main peak (%)	99.6	99.8	99.2	99.5
		HMW (%)	0.4	0.2	0.8	0.5
	CE-SDS (NR)	Main peak (%)	95.2	98.1	96.1	98.6
		LMW (%)	4.9	2.0	3.9	1.5
Impurity	resDNA	ppb	1.2	<1.1	< 0.4	< 1.1
	HCP	ppm	21.4	5.7	22.5	3.3
	resProA	ppm	0.3	0.0	0.2	< 0.0

CHT clearance of HCP species implicated in PS-80 degradation was also evaluated. PS-80 is a common excipient in the drug substance (DS) formulation to maintain biomolecule stability, and enzymatic activity of certain populations of HCPs can induce PS-80 degradation and thus impact product stability. The CHT load was spiked with PS-80, and PS-80 levels were assessed at timepoints up to 7 days. The CHT product pool demonstrated very minimal PS-80 degradation over time, in contrast to the CHT load pool, suggesting clearance of HCP species that may impact PS-80 stability (Fig. 3).

Figure 3.

PS-80 spiking assay of BsAb in-process product pools. PS-80 levels remain stable in the CHT eluate pool, suggesting robust clearance of potential high-risk HCPs following CHT chromatography. PS-80 levels were measured at Days 0, 3, 5 and 7 following spiking into the CHT load and eluate pools.

In parallel, in-process pools were assessed for specific HCP populations by mass spectrometry (MS) analysis. By MS, potential high-risk HCP species, including lipases, were significantly reduced following CHT chromatography. In addition to lipases, high levels of other HCP species in the product pool prior to CHT purification including clusterin, 78 kDa glucose-regulated protein, serine protease HTRA1 and ACTB were significantly reduced proceeding from Protein A to CHT chromatography. In particular, lipases, widely known to promote PS-80 degradation [6, 7], were  cleared to undetectable levels in the CHT eluate pool (Table 4). Clearance of problematic HCP species is consistent with minimal PS-80 degradation in the PS-80 spiking assessment in comparison with that observed in the CHT load.

Table 4.

HCP characterization by LC/MS. By LC/MS, approximate HCP levels in ppm were determined for in-process samples. Select HCP species that may impact PS-80 degradation, including lipase, are cleared to a large extent by CHT chromatography. Determined concentrations are approximate due to limitations of the “Hi-3” method (for example, serine protease HTRA1 was likely underestimated in CHT Load); however, lipases were  not detected after CHT chromatography purification, including in UFDF or DS samples (data not shown)

Another critical factor in developing a biologics manufacturing process is scalability of each step of the process. In a pilot run, the CHT column also demonstrated consistent, robust performance. SEC-HPLC and CE-SDS purity as well as process-related impurity levels of product pools were evaluated. A comparable profile (data not illustrated) and similar performance to the bench-scale run were attained, with a CHT step yield of 82% and overall process yield of 70% from the Protein A step to CHT. High SEC-monomer purity of 99.5% and high CE-SDS (NR) main peak purity of 98.6% were achieved following CHT (Table 3). In addition, HCP, resDNA and resProA impurities were cleared to well-below acceptable limits, 3 ppm, < 1 ppb and 0 ppm, respectively (Table 3). Particularly, there was further LMW removal by CHT (Tables 3, 5 and Fig. 4). By CE-SDS (NR) analysis, the additional %LMW species predominantly cleared by CHT included LC, heavy light (HL) species and heavy-heavy-light (HHL) chain variants, also illustrated in electropherograms of CHT load and eluate product pools (Fig. 4). Overall %LMW levels were reduced down to low levels (1.5%) in the final CHT pool (Tables 3 and 5). Moreover, CHT chromatography resulted in further reduction of LC, HL, HH and HHL  species (Table 5).

Table 5.

%LMW variant clearance by CHT. Further removal of light chain (LC), heavy chain (HC), heavy light (HL), heavy heavy (HH) and heavy heavy light (HHL) LMW variants is achieved proceeding from CHT load to CHT eluate. In particular, species in which an intermediate polishing step are not effective in clearing, including, LC and HHL variants, are further separated from the BsAb heterodimer following CHT

CE-SDS (NR) LMW species	CHT load	CHT eluate
LC	0.563	0.136
HC	0.035	0.039
HL	0.716	0.430
HH	0.292	0.167
HHL	2.306	0.792

Figure 4.

CE-SDS electropherograms of in-process product pools. CE-SDS (NR) analysis indicates further clearance of LMW product variants proceeding from CHT load to CHT eluate pool.

DISCUSSION

BsAbs present unique downstream purification challenges due to the panel of product-related variants that must be monitored and cleared to generate high-purity DS. Here, we demonstrate incorporation of a mixed-mode resin, CHT, in a BsAb purification process, that is robust in separation of product- and process-related impurities from the main heterodimer molecule. We reveal generation of high-purity DS using this process for an IgG1 BsAb format that retains Fc effector function.

CHT was incorporated as the polishing step in a BsAb purification process. Gradient methodology was used to establish an optimal combination of sodium phosphate and sodium chloride conditions in the wash and elution steps to achieve differential separation for this basic BsAb. Increased sodium chloride concentrations were required for product elution at phosphate concentrations <20 mM. This is consistent with contributions by both metal affinity and cation exchange interactions to attain this separation [12]. In conjunction with impurity profile analyses, there was optimal separation from product-related LMW variants and process-related impurities to achieve high product quality under these conditions. Another favorable characteristic of CHT is scalability and consistent generation of product pool with sufficient purity at pilot scale. In a pilot scale-up run, high product quality was achieved following the final CHT polishing step. HMW aggregates were reduced to final levels of <1%. Notably, LMW species were further reduced by CHT. In particular, LC, half molecules, HC heterodimer and HHL BsAb variants that were not separated from the BsAb heterodimer by other chromatography steps, including the intermediate CEX step in the BsAb process as indicated by higher levels of these variants in the CHT load pool, were cleared by CHT chromatography. Consistent with these data, hydroxyapatite promoted enhanced clearance of BsAb variants, including three-quarter species and those containing an extra LC [4, 11].

However, due to the diversity of BsAb formats and resulting variety of mispaired entities, further development work would be required to optimize polishing resins as well as wash and elution conditions to maximize yield and impurity separation for each type of molecule [9, 20–24]. Initial gradient experiments can reveal differential separation of impurities from the main BsAb heterodimer [9]. A linear salt gradient on CHT Type II separated aggregate impurity from the main peak for an IgG1 BsAb, resulting in 97% SEC purity [9]. BsAbs containing single-chain variable fragments (sc-Fvs) can be more prone to aggregation and thus may require more stringent wash and elution parameters to clear aggregates [25]. Other mixed-mode resins including Capto MMC and Toyopearl MX-Trp 650 M that employ hydrophobic and cation exchange mechanisms can also be effective in achieving high BsAb purity, but may only be suitable for particular BsAb molecules [4, 26–28]. Capto MMC reduced aggregate levels present in the affinity chromatography product pool to <1% for one BsAb, but levels remained higher at about 5% for a different BsAb [27]. Through a linear pH gradient on Capto MMC Impres, wash and elution conditions effective in separation of half antibody and homodimer impurity species for a BsAb molecule were established, but at the expense of lower yields [26]. Thus, polishing conditions for CHT and other mixed-mode resins need to be refined to maximize both BsAb purity and yield. Since CHT exhibited optimal impurity separation, consistent and scalable performance, as well as successful application to purification of other BsAbs, CHT was selected for this particular BsAb.

In addition to product-related impurities, CHT was effective in clearance of potential PS-80-degrading HCP populations. In a PS-80 spiking study, levels of PS-80 in the CHT eluate pool remained stable. Select HCP species implicated in PS-80 degradation [7], including lipases, were further cleared following CHT chromatography. Initial high levels of clusterin, 78 kDa glucose-regulated protein, serine protease HTRA1 and ACTB, species that have also been identified in mAb cell culture supernatants by proteomic analyses and associated with protein dynamics and apoptosis regulation [29], were significantly reduced following CHT. Although initial levels of some potential problematic HCPs were low, minimal abundance HCPs can retain the ability to promote enzymatic degradation of PS-80 [6, 8, 30]. An intermediate CEX step was insufficient to clear some problematic HCPs, as indicated by reduction in %PS-80 over time in the spiking assessment for the CHT load material. Thus, the BsAb purification process consisting of a mixed-mode step is ideal for generation of high purity BsAb heterodimer.

With BsAb structures advancing as pivotal immuno-oncology therapeutics, there is a need for a robust downstream process due to the complexity and unique panel of product-related variants associated with these molecules. Mixed-mode CHT chromatography plays a critical role in the downstream BsAb purification process by removing product-related impurities (HMW and LMW species) and process-related impurities (HCP, resDNA, resProA). CHT demonstrates robust performance, optimal scalability and effective separation of novel product-related variants associated with BsAbs as well as PS-80-degrading HCPs.

FUNDING

This study was supported by Bristol Myers Squibb.

CONFLICT OF INTEREST STATEMENT

J.A.W., J.Z., R.D., T.S., B.U., A.L., W.W., L.T. and S.G. are all employees of Bristol Myers Squibb.

DATA AVAILABILITY

The data underlying this article are available in the article.

ETHICS AND CONSENT STATEMENT

Not applicable.

ANIMAL RESEARCH STATEMENT

Not applicable.