Selection of bispecific antibodies with optimal developability using FcRn‑pH‑HPLC as an optimized FcRn affinity chromatography method

ABSTRACT

A challenge when developing therapeutic antibodies is the identification of candidates with favorable pharmacokinetics (PK) early in development. A key determinant of immunoglobulin (IgG) serum half‑life in vivo is the efficiency of pH-dependent binding to the neonatal Fc receptor (FcRn). Numerous studies have proposed techniques to assess FcRn binding of IgG-based therapeutics in vitro, enabling prediction of serum half-life prior to clinical assessment. FcRn high-performance liquid chromatography (HPLC) assays FcRn binding of therapeutic IgGs across a pH gradient, allowing the correlation of IgG column retention time to the half‑life of a therapeutic IgG in vivo. However, as FcRn retention time cannot be directly compared to an in vivo parameter, modifications to FcRn-HPLC are required to enable interpretation of the data within a physiological context, to provide more accurate estimations of serum half-life. This study presents an important modification to this method, FcRn-pH-HPLC, which reproducibly measures FcRn dissociation pH, allowing correlation with previously established half-lives of therapeutic antibodies. Furthermore, the influence of incorporating various antibody modifications, binding modules, and their orientations within IgGs and bispecifics on FcRn dissociation pH was evaluated using antibodies from the redirected optimized cell killing (ROCK®) platform. Target and effector antigen-binding domain sequences, their presentation format and orientation within a bispecific antibody alter FcRn retention; tested Fc domain modifications and incorporating stabilizing disulfide bonds had minimal effect. This study may inform the generation of mono-, bi- and multi-specific antibodies with tailored half-lives based on FcRn binding properties in vitro, to differentiate antibody-based therapeutic candidates with optimal developability.

Video Abstract

Read the transcript

Watch the video on Vimeo

© 2023 Affimed GmbH. Published with license by Taylor&Francis Group, LLC

KEYWORDS:

• antibody therapies
• bispecific
• FcRn-Ph-HPLC
• ICE®
• immunotherapy
• multispecific
• ROCK® platform
• therapeutic IgG

CLASSIFICATIONS:

• Antibody analytics
• Antibody engineering
• Antibody generation
• Antibody screening
• Antibody structure
• Bispecific
• Chromatography
• FcRn
• Pharmacokinetics

Introduction

Monoclonal antibody (mAb)-based therapeutics are a vital treatment strategy for patients with various diseases.¹ Among these, bispecific antibodies (bsAbs) are a growing class of antitumor therapeutics that can target two antigens simultaneously, enabling novel mechanisms of action inaccessible to standard mAbs.² These novel mechanisms include the recruitment of innate and adaptive immune cells to tumor sites for the induction of anti-tumor antibody-dependent cellular cytotoxicity (ADCC),^3–5 and dual targeting of tumor-associated antigens.⁶ The efficacy, method of administration and dosing-regimen of IgG-based antibody therapeutics are heavily reliant on their serum half-life and their pharmacokinetic (PK) properties.⁷ A persistent challenge in the preclinical characterization of IgG therapeutics is the identification of drug candidates that are most likely to exhibit the optimal PK required to elicit the greatest response and most convenient dosing strategy in patients with a particular disease.⁸ Methods to accurately predict and monitor half-life and other PK properties are needed to identify IgG therapeutics with the greatest developability and feasibility for clinical use early in the drug discovery process.^7,9

The serum half-life of IgGs in vivo is largely established through their binding to the neonatal fragment crystallizable receptor (FcRn).¹⁰ FcRn is a heterodimeric, non-classical Fc-gamma receptor (FcγR) consisting of a 40 kDa α heavy chain non-covalently associated with a β₂microglobulin light chain, which shares structural similarities with major histocompatibility class I molecules.^11,12 Although the receptor lacks a direct role in immune responses, FcRn is responsible for the rescue and recycling of serum IgG pinocytosed from blood serum into vascular endothelial cells and antigen-presenting cells such as macrophages, dendritic cells, and subsets of B cells.^13,14 FcRn exhibits a predominantly intracellular localization within the mildly acidic environment of early endosomes,¹⁵ where it associates at a stoichiometry of 2:1 with an internalized IgG molecule.¹⁶ This association is highly pH-dependent, requiring the protonation at around pH 6.0 of two histidine residues (His310 and His435) within the CH2-CH3 interface of the IgG Fc-domains, which can then interact with the anionic residues Glu115 and Asp130 on the FcRn α chain.^17,18 Following association, FcRn/IgG complexes are sorted into recycling endosomes, followed by further trafficking through various exocytic pathways for release into serum.¹⁹ In contrast, unbound IgG is trafficked to lysosomes for degradation.²⁰ Following exocytosis of FcRn/IgG complexes, pH-dependent dissociation at the physiological pH of 7.4 occurs due to loss of protonation of the Fc-domain His310 and His435 residues, enabling re-release of IgG back into serum.²¹ Failure of IgG to dissociate from FcRn at this stage results in expedited clearance of the antibody.^10,12 As such, the efficiency with which an IgG-based antibody binds FcRn at pH 6.0, as well as the efficiency of dissociation at pH 7.4, are key determinants of IgG half-life.^20,22 Physiochemical properties of the IgG, such as the charge distribution of the variable fragment antigen-binding (Fab) domains²³ and the isoelectric point (pI) of individual IgG domains,^9,22 post-translational modifications, such as oxidation,⁹ and aggregation⁹ have also been found to influence IgG binding to FcRn, and thus play a role in determining half-life and other PK properties. The influence of other posttranslational modifications, namely glycosylation, on PK properties, however, remains controversial,^24,25 with some studies reporting that no differences in PK are observed by the same antibody in its glycosylated and deglycosylated forms in cynomolgus monkeys²⁶, whereas other studies suggest that antibodies with certain glycans incorporated, such as certain high-mannose-type glycans, do exhibit differences in clearance in humans compared with antibodies that incorporate different glycans.^27,28 Mechanistically, the aforementioned physicochemical properties, charge distribution in different antibody modules, pI, as well as post-translational modifications and aggregation, can induce conformational changes in the Fc portion, interfering with the binding between the Fc domain and FcRn.²⁹

It has previously been hypothesized that in vitro FcRn binding assays could be used to estimate PK properties of therapeutic IgGs in vivo.⁹ Of the available techniques,^30–35 surface plasmon resonance assays are most commonly used,^{30–32,36,37} but they have key limitations, including high complexity and inconsistency in correlating in vitro FcRn binding to IgG serum half-life in vivo.^33,38,39

A human FcRn affinity liquid chromatography method that assesses the binding of IgG to FcRn under near-physiological conditions has been developed to more accurately predict the PK properties of IgG-based therapeutics.^9,40 In this assay, human FcRn molecules are immobilized on beads on a high-performance liquid chromatography (HPLC) column, with a linear pH gradient of pH 5.5 to 8.8 established across its length. Antibodies are then injected at low pH and elution assessed across the pH gradient. This method has previously been used to demonstrate a clear correlation between FcRn binding of IgGs containing Fc-domain mutations and their respective PK properties in vivo, showing that this technique may be used to differentiate therapeutic IgGs with optimal developability in the early stages of drug development. Furthermore, coupling of this technique,⁴⁰ and other similar column-based FcRn-affinity techniques such as capillary electrophoresis,⁴¹ to mass spectrometry has enabled greater elucidation of the impact of IgG modifications on FcRn binding characteristics, further demonstrating the utility of these techniques to optimize therapeutic IgGs.

The established redirected optimized cell killing (ROCK®) platform comprises a plethora of known and novel bi- and multi-specific antibody modules and formats, including IgG-like bsAbs that can be used to simultaneously engage CD16A on natural killer (NK) cells and macrophages, and a tumor cell-surface antigen.³ Simultaneous engagement of CD16A and a tumor cell surface antigen results in the activation and re-direction of NK cells and macrophages to tumor cells, stimulating anti‑tumor ADCC and antibody-dependent cellular phagocytosis, respectively.³ From this platform, the tetravalent, bispecific innate cell engager (ICE®) antibody therapies AFM13 and AFM24 are currently in clinical development for CD30⁺ lymphomas and EGFR⁺ solid tumors, respectively.^42–46 It was, therefore, of interest to characterize the FcRn binding capabilities of different ROCK® platform-derived antibody modules and formats to help facilitate the design and development of novel antibody therapies with optimal serum half-lives.

In this study, we present an advanced FcRn affinity HPLC method that combines a pH monitor with the previously described human FcRn affinity HPLC (FcRn-pH-HPLC). This method allowed the assessment of the pH at which IgG-based antibodies dissociated from human FcRn reliably and reproducibly, enabling assessment of FcRn binding properties in vitro, and facilitating direct comparison with the optimum FcRn dissociation pH exhibited by physiological IgG in vivo. FcRn‑pH‑HPLC was then used to characterize the FcRn dissociation pH of a range of antibodies with mutations commonly utilized in the engineering of therapeutic antibodies, and to systematically investigate how different ROCK® modules/components, as well as their stoichiometry and orientation, contribute to the FcRn dissociation pH of IgG-based bsAbs. This study demonstrates that the FcRn‑pH-HPLC method enables accurate identification of the pH at which novel therapeutic antibody candidates dissociate from FcRn, allowing prediction of potential serum half-lives of candidate therapeutics to aid the development of novel IgG-based antibody therapies with tailored PK profiles.

Results

The FcRn‑Ph-HPLC-HPLC method allows consistent evaluation of the dissociation pH of antibody constructs from FcRn

To measure the pH at which a therapeutic IgG antibody dissociates from FcRn, thus providing a direct comparison with the optimal FcRn dissociation pH of physiological IgG, FcRn affinity HPLC was combined with a pH monitor (FcRn-pH-HPLC). To assess the reliability of FcRn-pH-HPLC to establish the FcRn dissociation pH consistently, the technique was performed using NIST mAb, a reference mAb (reference material RM8671) commonly used for assessing the performance of techniques measuring the properties of mAbs.⁴⁷ Briefly, a pH monitor was subjected to two-point calibration prior to NIST mAb injection onto an HPLC column containing immobilized, unmodified, wild-type (WT) human FcRn. Elution was then performed as previously described,⁹ and the pH of molecules dissociating from FcRn determined by continuous measurement of pH and absorbance at 280 nm. To ensure variability was fully assessed, two different lots of NIST mAb were used, across 6 d, and 15 independent runs performed in triplicates; no more than 90–100 injections were performed using the same FcRn HPLC column to minimize the effects of the degradation of immobilized active FcRn.

FcRn‑pH-HPLC reliably and reproducibly measured FcRn dissociation pH of NIST mAb, as minimal variance was observed across the samples tested (Figure 1A). The mean (standard deviation, SD) dissociation pH of NIST mAb from FcRn was pH 7.216 (±0.052), with a median (range) dissociation pH of 7.206 (7.145–7.315). Furthermore, FcRn dissociation of NIST mAb was compared to that of human normal immunoglobulin (Cutaquig®). While the monoclonal NIST mAb dissociates as a single peak from FcRn, polyclonal immunoglobulin dissociated with a major and minor peak at pH 7.254 and 7.497, respectively (Supplementary Figure S1) consistent with previous studies measuring FcRn dissociation using retention time.⁴⁸ As such, FcRn-pH-HPLC resolves different antibody species and isotypes within a polyclonal mixture, suggesting the potential to interrogate pH-dependent elution profiles of human serum IgGs.

Figure 1. FcRn-Ph-HPLC is capable of reliably assessing the FcRn dissociation pH of mAbs, which can be correlated with half-life established in vivo in humans.

(a) Bar graph depicting a single bar, with error bars and 15 individual data points, showing the FcRn dissociation pH of NIST mAb control antibodies. (b) Bar graph of FcRn dissociation pH values for cusatuzumab, cetuximab, margetuximab, durvalumab, rituximab, adalimumab, and briakinumab. (c) Bar graph showing differences in FcRn dissociation pH for cusatuzumab, cetuximab, margetuximab, durvalumab, rituximab, adalimumab, and briakinumab from physiological pH (7.4), with data points/ranges depicting the mean of the previously established, human half-life values in humans for each antibody, as reported in the prescribing information or a representative study.

Notes: (a) FcRn dissociation pH of NIST mAb control antibody as assessed by FcRn-pH-HPLC. Mean ± SD of pH values are shown from 15 independent experiments. (b) Means ± SD are depicted for each antibody, calculated from three independent experiments. (c) Bars: Mean difference in measured FcRn dissociation pH from pH of physiological IgG in vivo (pH 7.4) as established by FcRn-pH-HPLC. Values were calculated using the following: FcRn dissociation pH of physiological IgG (pH 7.4) ˗ mean FcRn dissociation pH of the indicated antibody. Mean ± SD are shown for each antibody. Data points/ranges (in green): In vivo half-lives in human of the indicated therapeutic mAbs, as established previously in clinical studies and presented in the prescribing information or a representative clinical study for each antibody.^49-55

To determine whether the FcRn dissociation pH established using this technique could be correlated with IgG serum half-life, seven representative therapeutic mAbs with well‑characterized PK profiles in humans were then examined using FcRn-pH-HPLC, as before (Figure 1B). As observed for the relative FcRn column retention time used in previous studies,⁹ the FcRn dissociation pH values obtained demonstrated a correlation with reported half-lives of therapeutic mAbs (Figure 1C). Rituximab, adalimumab, margetuximab, and durvalumab, which have longer observed half-lives in vivo (18–32 , 10–20 , 19.2, and 21 d, respectively)^49–52 relative to the other tested therapeutic mAbs exhibited the mean [SD] FcRn dissociation pH values (7.431 [±0.053], 7.389 [±0.049], 7.339 [±0.049], and 7.277 [±0.048], respectively) most similar to physiological IgG (pH 7.4); the optimal FcRn dissociation pH. In contrast, cetuximab, cusatuzumab and briakinumab, which have the shortest observed half-lives in vivo (reported means: 3.4–4.7, 6.1–10.4, and 9.0 d, respectively),^53–55 had mean [SD] FcRn dissociation pH values (7.251 [±0.055], 7.115 [±0.041] and 7.813 [±0.047], respectively) most dissimilar to pH 7.4. As such, mAbs that exhibited dissociation from FcRn at pH values most similar to that of physiological IgG, pH 7.4, appeared to exhibit more favorable previously reported serum half-lives. Moreover, antibodies examined in this study with previously reported half-lives ≥19.2 d exhibited an FcRn dissociation pH between 7.277 and 7.431, suggesting that this could potentially be a favorable range in which to tailor therapeutic mAbs for which a long half-life is desired.

Fc domain modifications commonly used to generate therapeutic IgGs do not affect FcRn retention

Having established that FcRn-pH-HPLC could produce FcRn retention data which correlated well with in vivo PK properties of therapeutic mAbs, the impact of common Fc and antigen-binding domain modifications used in the development of therapeutic IgGs on FcRn retention was investigated. Firstly, representative modifications of Fc domains, which enable modulation of antibody effector functions and play an important role in the production of bi- and multi-specific formats, were explored. CD19-specific IgGs (MOR208) with WT Fc, and Fc domains modified with either Fc‑silencing mutations in the CH2 (N297G; FEAG: L234F, L235E, D265A, N297G; LALAPG: L234A, L235A, P329G), Fc‑enhancing mutations in the CH2 (S239D, I332E), or a Knobs‑into-Holes (KiH) knob mutation (T366W) combined with KiH hole mutations (T366S, L368A, Y407V) in the CH3, were subjected to FcRn-pH-HPLC, and FcRn dissociation pH was compared to WT.

As the mutations tested are located in the Hinge-CH2 region or are buried within the CH3‑CH3 region, and thus outside of the main FcRn/Fc interface in the CH3 sub-domain⁵⁶ (Figure 2A), the FcRn binding region is largely unaltered with the mutations described above. As such, none of the tested anti-CD19 IgG variants exhibited significantly different FcRn dissociation pH values compared to WT, or in comparison among the Fc-modified variants (p ≥ 0.05), with mean (SD) FcRn dissociation pH values ranging from 7.235 (±0.048) to 7.301 (±0.028) (Figure 2B).

Figure 2. Examples of common Fc domain mutations used in the engineering of therapeutic IgGs do not significantly influence FcRn dissociation pH.

(a) Two IgG images describing the mutations introduced to the Fc domain to assess the influence of Fc modifications on FcRn dissociation pH. (b) Bar graph depicting the FcRn dissociation pH values observed for MOR208 IgG molecules with WT, Fc-silenced, Fc-enhanced, or KiH Fc domain modifications incorporated.

Notes: (a) Schematic IgG showing the location/position of Fc domain mutations, designed to silence effector function (FEAG, LALAPG, N297G), enhance effector function (S239D, I332E), or allow heterodimeric antibody assembly via KiH mutations. (b) FcRn dissociation pH of WT MOR208 mAb, and MOR208 mAbs with Fc-silencing, Fc-enhancing, or KiH mutations, as indicated. Mean ± SD is depicted for each antibody, calculated from three independent experiments.

Use of effector domains specific to a common antigen, but with different antigen-binding sequences, significantly alters FcRn retention in vitro

Altering the sequence of antigen-binding domains to tailor the binding affinity of therapeutic antibodies can also alter biochemical properties of the domain, such as the pI, which has previously been shown to affect FcRn retention. To assess whether this was the case for effector domains derived from the ROCK® platform, three proprietary antiCD16A effector antigen-binding domains presented in Fab format, previously generated to have a low, medium, and high affinity for CD16A and incorporated within a standard IgG, were subjected to FcRn-pH-HPLC (Figure 3a). The low-, medium-, and high-affinity domains had pI values of 7.78, 7.66, and 7.88, respectively.

Figure 3. Incorporating different antigen-binding domain sequences to a common antigen significantly influences the FcRn dissociation pH.

(a) Schematic depicting an IgG molecule with the CD16A effector antigen-binding domains highlighted, and text describing whether these had low, medium, or high affinity for CD16A, and that the Fc domains are identical. (b) Bar graph depicting significant differences in FcRn dissociation pH of IgGs incorporating either low, medium, or high affinity CD16A effector binding domains.

Notes: (a) Schematic showing standard IgG antibody format into which anti-CD16A effector antigen-binding domains with different sequences and affinities were incorporated. (b) FcRn dissociation pH values obtained for antibodies with low-, medium-, or high-affinity CD16A sequences incorporated in N-terminal Fab domains. Mean ± SD is presented for each antibody variant, calculated from two–three independent experiments. Statistical significance (p ≤ 0.05) was determined by unpaired T-test.

The choice of sequence used to generate the anti-CD16A effector domains significantly influenced the FcRn dissociation pH (Figure 3b). The low-, medium-, and high-affinity domains exhibited mean (SD) elution pHs of 7.336 (±0.051), 7.434 (±0.024), and 7.195 (±0.037), respectively. This suggested that the biochemical properties, attributed to the differing sequences, such as pI, may alter FcRn binding, and thus potentially the PK properties of the IgGs.

Increasing the number of CD16A-binding effector domains fused to the Fc C-terminus in multivalent bsAbs results in an increase in FcRn retention

Having established that the choice of effector antigen-binding domain significantly influenced the FcRn retention of IgGs, we then assessed how fusing these domains to the C-terminus of an IgG, thus generating a multivalent bsAbs, would influence FcRn binding. Furthermore, the effect of increasing the number of these domains was also examined. As such, four ROCK® platform-derived bsAbs with identical Fc and N-terminal Fab domains, but with varying numbers of identical C-terminal anti-CD16A domains, were assessed using FcRn-pH-HPLC as before (Figure 4). To assess the effect of adding a greater number of C-terminally fused scFv modules, an IgG antibody with no fused C‑terminal CD16A domains (IgG) was compared to a tetravalent 2Fab‑Fc‑2scFv bsAb and a hexavalent bsAb with two C‑terminally fused single-chain diabody (scDb) modules (2Fab-Fc-2scDb). Furthermore, to assess the impact of increasing the number of scDb modules, the tetravalent bsAb 2Fab-Fc-1scDb was compared to the hexavalent 2Fab-Fc-2scDb bsAb (Figure 4A).

Figure 4. The number of antigen-binding domains presented in scFv or scDb significantly influences FcRn dissociation pH.

(a) Images of IgG, 2-Fab-Fc-2scFv, 2Fab-Fc-1scDb, and 2Fab-Fc-2scDb antibodies, with the numbers of C-terminal effector domains incorporated highlighted. Text highlights that all antibodies shown had identical Fab and Fc domains. (b and c) Two bar graphs depicting significant differences between the FcRn dissociation pH values of IgG, 2Fab-Fc-2scFv, and 2Fab-Fc-2scDb (left graph), and between 2Fab-Fc-1scDb and 2Fab-Fc-2scDb (right graph).

Notes: (a) Schematic figures depicting the two pairs of bsAbs used to investigate the effect of antigen-binding domain number on FcRn dissociation pH. (B and C) FcRn dissociation pH established by FcRn-pH-HPLC for (b) different numbers of C-terminally fused CD16A domains (2× CD16A binding moieties in 2× scFv vs. 4× CD16A binding moieties in 2× scDb) and (c) different numbers of C-terminally fused CD16A binding domains in scDb modules; mean ± SD are shown, calculated from three independent experiments. Statistical significance (p ≤ 0.05) was determined by unpaired T-test for each comparison. Note: the datasets for 2Fab-Fc-2scDb bsAbs in B and C were performed with two different versions of this antibody, incorporating two different target antigen‑binding domains. As such, these datasets cannot be consolidated into a single dataset.

Fusion of anti-CD16A domains in scFv format to the C-terminus of an IgG to generate a multivalent bsAb significantly increased the FcRn dissociation pH, with the 2Fab‑Fc‑2scFv exhibiting mean (SD) dissociation at pH 7.525 (±0.058); a mean increase of 0.325 pH units over that of the IgG (pH 7.200 [±0.054]). Furthermore, increasing the number of antigen-binding domains fused to the C-terminus, in scFv or scDb format, significantly increased the FcRn dissociation pH. Fusion of four anti‑CD16A domains, presented as two scDb modules, increased the mean (SD) FcRn dissociation pH from 7.200 (±0.054) to 7.687 (±0.058), a mean increase of 0.487 pH units compared to IgG and 0.162 above that of the 2Fab-Fc-2scFv (Figure 4B). Similarly, when increasing the number of antigen-binding domains incorporated in scDb format from two (2Fab-Fc-1scDb) to four (2Fab-Fc-2scDb) (Figure 4C), adding an additional scDb fusion module increased the mean (SD) FcRn dissociation pH from 7.468 (±0.043) to 7.725 (±0.047). The fusion of C‑terminal effector domains, and the numbers of these fused, are, therefore, potentially key contributors to the overall FcRn retention properties of therapeutic antibodies.

To investigate potential contribution of individual modules of bsAbs and exclude nonspecific interactions between the extra domains and the column matrix, we examined the FcRn dissociation pH values of isolated modules comprising tested bsAbs, including scFv, scDb, Fab, and a human IgG1 Fc, using FcRn-pH-HPLC, and compared the results to the those of the whole bsAb or IgG (Supplementary Figure S2). The scFv, scDb, and Fab fragments showed no or minimal interaction with the FcRn column, eluting prior to the pH gradient being applied (Supplementary Figure S2). However, when the same modules were fused to the IgG in our bsAb platform, such as in 2Fab-Fc-2scFv (Supplementary Figure S2a) or 2Fab-Fc-1scDb (Supplementary Figure S2b), retention time and FcRn dissociation pH were increased beyond the values observed for the IgG construct without such a C-terminal fusion. Similarly, comparison of the FcRn dissociation pH and column retention time of cetuximab with its isolated Fab module showed no or minimal binding of the Fab module alone, whereas fusion of the Fab module to the IgG Fc domain to form cetuximab decreased the observed FcRn dissociation pH and column retention time below that of the IgG Fc portion alone (Supplementary Figure S2c).

Non-specific interactions of antibody constructs lacking Fc domains were, therefore, excluded; however, fusion of such modules to Fc portions influences FcRn interaction and can be used to modulate FcRn binding and serum half-life in multispecific antibody formats.

Utilizing different target antigen-binding sequences induces large variations in the FcRn retention properties of bsAbs in vitro

Incorporating antigen-binding domains specific to different antigens within standard monoclonal therapeutic IgGs with similar or identical Fc domains has previously been shown to affect FcRn binding, and consequently their predicted or observed serum half-lives, based on differing properties such as the pI of their respective complementarity-determining regions (CDRs).²² Thus, continuing with the evaluation of the FcRn binding properties of bsAbs, the effect of incorporating N-terminal target antigen-binding domains, specific to different antigens, into bsAbs was then examined, alongside the possibility of the FcRn dissociation pH relating to the pI of the bsAbs (Figure 5). To do this, 2Fab-Fc-2scFv bsAbs with identical Fc and C-terminally fused antiCD16A single-chain fragment variable (scFv) modules but Fab modules specific to different target antigens (CD20, CD19, RSV, CD123, CLL‑1, CD30, AMHRII, CXCR4, or PSCA) were subjected to FcRn‑pH‑HPLC to establish the FcRn dissociation pH. As for standard monoclonal therapeutic IgGs, the incorporation of different target antigen-binding domains in the Fab module in a bsAb influenced the mean FcRn dissociation pH (range: 7.317 to 7.560), with bsAbs exhibiting a wide variation in FcRn dissociation pH based on the target antigen-binding domain incorporated. Overall, no clear relationship was observed between the pI and the FcRn dissociation pH of the different bsAbs. However, within the pI range of 8.0–8.5, FcRn dissociation pH does appear to generally increase as pI increases, suggesting there may be some correlation between these two parameters within this range.

Figure 5. Incorporation of different target domain sequences specific to different antigens within N-terminal Fab modules in bsAbs influences FcRn dissociation pH.

Schematic figure (a) depicts a 2Fab-Fc-2scFv bispecific antibody with the N-terminal target antigen-binding domains highlighted, and text describing these domains as being the independent variable in this experiment. Additional text denotes that the Fc and scFv parts of all antibodies tested in this experiment were identical. A correlation plot (B) presents FcRn dissociation pH plotted against the isoelectric point of the target antigen-binding domain incorporated.

Notes: (a) FcRn dissociation pH of bsAbs containing different N-terminal target domain sequences in Fab modules. (b) Isoelectric point/FcRn dissociation pH correlation plot.

The introduction of stabilizing disulfide bridges within scFv domains does not significantly influence FcRn retention

Another common modification of bsAbs is the incorporation of disulfide bridges (DSB) within scFv module, to stabilize the structures and prevent aggregation as the concentration increases. To determine whether scFv DSB influenced FcRn dissociation pH, Arg‑44 and Ser‑105 in the variable heavy and variable light regions of the scFv domains in a 2Fab-Fc-2scFv tetravalent bsAb were mutated to Cys residues, to generate intramolecular stabilizing DSBs as previously described.⁵⁷ These were then examined using FcRn-pH-HPLC, and FcRn dissociation pH compared to a 2Fab-Fc-2scFv bsAb without stabilizing DSBs (Figure 6A).

Figure 6. Incorporation of stabilizing DSB within scFv modules of bsAb does not significantly influence FcRn dissociation pH.

(a) Two 2Fab-Fc-scFv bispecific antibody schematics with the C-terminal effector antigen-binding domains highlighted. The left antibody contains stabilizing disulfide bonds, the right antibody does not. Additional text highlights that the N-terminal target antigen-binding domains, and the Fc portions of the antibody are identical. (b) Bar graph depicting the FcRn dissociation pH of the 2Fab-Fc-2scFv antibodies with and without stabilizing disulfide bonds.

Notes: (a) Schematic depicting the tetravalent, bsAb formats (2Fab-Fc-2scFv) used to investigate the FcRn dissociation pH of antibodies with (rightleft) and without (leftright) stabilizing DSB within their C-terminal scFv modules. (b) FcRn dissociation pH values obtained for antibodies with and without stabilizing DSB incorporated in C-terminal scFv modules. Mean ± SD is presented for each antibody variant, calculated from three independent experiments. Statistical significance (p ≤ 0.05) was determined by unpaired T-test.

Stabilization of the C-terminal scFv modules through the introduction of stabilizing DSBs did not significantly alter the FcRn dissociation pH (Figure 6B); the bsAbs with and without stabilizing DSB bridges yielded mean (SD) dissociation pH values of 7.501 (±0.055) and 7.463 (±0.056), respectively.

Presenting the antigen-binding domain in Fab or single-chain diabody format significantly affects FcRn retention in bsAbs derived from the ROCK® platform

After establishing the influence of common antibody modifications on FcRn retention, a systematic evaluation of the effect of different types of antibody modules, and the orientation of these modules within bsAbs on FcRn dissociation pH was conducted utilizing different components of bsAb formats from the ROCK® platform. First, the effect of incorporating the antigen-binding domain as different modules on FcRn dissociation pH was evaluated. Six tetravalent, bispecific antibodies derived from the ROCK® platform were used to compare the effect of incorporating an antigen-binding domain, with identical antigen-specific sequences and valency, in Fab, scFv, or scDb module on FcRn dissociation pH using FcRn-pH-HPLC (Figure 7A).

Figure 7. The structure in which antigen-binding domains are presented significantly influences FcRn dissociation pH.

(a) Schematics depicting three pairs of tetravalent, bispecific antibodies: Pair 1 = 2scFv-Fc-2scFv and 2Fab-Fc-2scFv; pair 2 = 1stgFab-1scDb-Fc and 2stgFab-Fc; pair 3 = 2Fab-Fc-2scFv and 2Fab-Fc-1scDb. Differences in how the target and effector antigen-binding domains are presented between the antibodies within each pair are highlighted; additional text depicts portions of the antibodies within each pair that are identical. (b) Bar graph depicting FcRn dissociation pH values observed for 2scFv-Fc-2scFv and 2Fab-Fc-2scFv. (c) Bar graph depicting FcRn dissociation pH values observed for 1stgFab-1scDb-Fc and 2stgFab-Fc. (d) Bar graph depicting FcRn dissociation pH values observed for 2Fab-Fc-2scFv and 2Fab-Fc-1scDb.

Notes: (a) Schematic figures depicting the three pairs of tetravalent, bsAbs used to investigate the effect of antigen-binding domain structure on FcRn dissociation pH. (B–D) FcRn dissociation pH established by FcRn-pH-HPLC for (b) different N-terminal fused target domain presenting modules (scFv or Fab), (c) different N-terminal fused CD16A effector domain presenting modules (scDb or stgFab), and (d) different CD16A C-terminal fused effector domain presenting modules (2× scFv or 1× scDb); mean ± SD are shown, calculated from three independent experiments. Statistical significance (p ≤ 0.05) was determined by unpaired T-test for each comparison.

When comparing antibodies with identical Fc domains and C-terminal anti-CD16A scFv modules, but with the N-terminal target antigen-binding domains presented as either scFv modules (2scFv-Fc-2scFv) or Fab modules (2Fab-Fc-2scFv), there was no significant difference in the FcRn dissociation pH, with mean (SD) dissociation values of the 2scFv-Fc-2scFv and 2Fab-Fc-2scFv bsAbs being pH 7.593 (±0.027) and 7.544 (±0.038), respectively. This suggests these could be used interchangeably on the N-terminus (Figure 7B). In contrast, incorporating an N-terminal CD16A domain in a staggered Fab (stgFab) structure (2stgFab-Fc) lowered the FcRn dissociation pH closer to pH 7.4 when compared to utilizing an scDb (1stgFab-1scDb-Fc) (Figure 7C); mean (SD) FcRn dissociation pH values were 7.560 (±0.032) and 7.352 (±0.034) for the 1stgFab-1scDb-Fc and 2stgFab-Fc antibodies, respectively. Comparing the 2Fab-Fc-2scFv bsAb with another bsAb having identical Fc and N-terminal Fab module, but with its two anti-CD16A binding domains in a single scDb module C-terminally fused to only one of the heavy chains (2Fab-Fc-1scDb), did not yield a significantly different FcRn dissociation pH (7.544 [±0.038] and 7.468 [±0.043], respectively), suggesting that these modules could be used interchangeably on the C-terminus (Figure 7D).

Re-orientating antigen-binding domains within a bsAb results in significant differences in FcRn retention

After establishing that the choice of module in which an antigen-binding domain was presented can affect the FcRn dissociation pH, the effect of changing the location of these domains and modules within a bsAb was subsequently examined. Four tetravalent, bsAbs with identical specificities, Fc domains and antigen‑binding domain sequences, but with the antigen-binding domains presented in different modules and different location within the antibodies, were derived from the ROCK platform (Figure 8A) and subjected to FcRn-pH-HPLC.

Figure 8. The orientation of antigen-binding domains within a bsAb significantly influences FcRn dissociation pH.

Notes: (a and b) Schematic figures depicting the two pairs of tetravalent, bsAbs used to investigate the effect of antigen-binding domain orientation on FcRn dissociation pH. Black boxes indicate the modules which have been re-orientated within the two pairs of bsAbs. (c and d) FcRn dissociation pH established by FcRn-pH-HPLC for (c) swapped N- and C-terminally fused binding domains (d) re-oriented Fab and scDb domains; mean ± SD are shown, calculated from three independent experiments. Statistical significance (p ≤ 0.05) was determined by unpaired T-test for each comparison.

Read the detailed description of this figure

(a) Images depicting a pair of 2Fab-Fc-2scFv bispecific antibodies, with the N-terminal (orange) and C-terminal (blue) antigen-binding domains highlighted. The image on the left depicts the target domains on the N-terminus and the effector domains on the C-terminus (2FabTarget-Fc-2scFvEffector); the image on the right depicts the effector domains on the N-terminus and the target domains on the C-terminus (2FabEffector-Fc-2scFvTarget). (b) Images depicting a pair of antibodies. The left image depicts an antibody with a staggered Fab and scDb module on the N-terminus (1stgFab-1scDb-Fc); the image on the right has two Fab modules on the N-terminus, and a scDb module on the C-terminus (2Fab-Fc-1scDb). (c) Bar graph depicting FcRn dissociation pH values observed for 2FabTarget-Fc-2scFvEffector and 2FabEffector-Fc-2scFvTarget. (d) Bar graph depicting FcRn dissociation pH values observed for 1stgFab-1scDb-Fc and 2Fab-Fc-1scDb.

BsAbs in which the target and CD16A antigen-binding domains had been swapped at the N- and C-termini were compared, i.e., a bsAb with target binding domains in Fab modules at the N-terminus and CD16A antigen-binding-domains in scFv modules at the C-terminus (2Fab^Target‑Fc‑2scFv^Effector) was compared to a bsAb in which the CD16A antigen-binding domains were incorporated into Fab modules on the N-terminus, and the target antigen-binding domains presented in scFv modules on the C-terminus (2Fab^Effector-Fc-2scFv^Target). Re-orientation of the target and effector domains significantly altered the FcRn dissociation pH (Figure 8b). Mean (SD) FcRn dissociation pH values were 7.544 (±0.038) and 7.420 (±0.042) for 2Fab^Target-Fc-2scFv^Effector and 2Fab^Effector-Fc-2scFv^Target, respectively. Presenting the CD16A domains on the N-terminus in the Fab module yielded the most similar dissociation pH values to physiological IgGs (pH 7.4).

To explore the impact of changing the location of the modules, bsAbs with CD16A antigen-binding domains incorporated into an scDb module, but with two target antigen-binding domains on the N‑terminus were compared. These antigen-binding domains were incorporated either as a stgFab on chain 1 of the bsAb with the scDb located on the N‑terminus of chain 2 (1stgFab-1scDb-Fc), or with the stgFab split into two Fab domains, with a Fab domain on each chain (N-terminus) and the scDb fused to the C‑terminus of the Fc domain chain 2 (2Fab-Fc-1scDb). Re‑orientation of Fab and scDb modules resulted in a significant difference in observed FcRn dissociation pH (Figure 8C). Mean (SD) values observed for 2Fab-Fc-1scDb and 1stgFab-1scDb-Fc were 7.468 (±0.043) and 7.560 (±0.032), respectively, with the 2Fab-Fc-1scDb exhibiting the dissociation pH most similar to physiological IgG of the two bsAbs.

Discussion

There is growing recognition that antibody-based therapeutics are a vital treatment strategy in many diseases and the advent of promising bsAbs with novel mechanisms of action may further improve responses. ^1,58–60 The ability to identify therapies with the greatest developability early in development is crucial to maximize the chances of clinical success.⁶¹ Optimal FcRn binding is recognized as a key determinant of half‑life for both physiological IgG and IgG-based antibody therapeutics,⁹ and various in vitro assays have sought to utilize recombinant human FcRn/IgG-based therapeutic binding to predict in vivo PK properties.^30–35 The development of FcRn‑HPLC has allowed a correlation between the output of an in vitro assay and in vivo half-life, representing a major step toward identifying and optimizing IgG‑based therapeutics with favorable PK properties early in preclinical development.⁹ Here, we present a further important modification to this technique, allowing analysis of the FcRn dissociation pH, facilitating the interpretation and classification of FcRn‑HPLC in a more physiological context.

In this study, combining FcRn-HPLC with an additional pH monitor enabled direct comparison between the dissociation of an IgG-based therapeutic from HPLC column-bound human FcRn, and physiological dissociation pH of an IgG from FcRn in vivo. FcRn-pH-HPLC can establish the FcRn dissociation pH of a range of well-characterized therapeutic mAbs, reliably and reproducibly, and the values obtained correlated with their previously established half-lives from studies in humans. Furthermore, this technique was proven capable of assessing the FcRn binding properties of both mAbs and bsAbs, and how these properties were influenced through incorporation of Fc modifications, disulfide bonds, different antigen-binding domains and modules of presentation, as well as their different orientations and location within a molecule. Data presented here, therefore, supports an alternative, extended method for FcRn-HPLC, allowing interpretation of the results obtained within a physiological context. Learnings from this study might be used to generate mAbs, bi- and multispecific antibodies with tailored PK properties based on FcRn dissociation properties.

Previous studies that incorporated FcRn‑HPLC used relative column retention time to assess the FcRn binding properties of IgG-based therapeutics.^9,62 This technique has been shown to be a robust method of predicting in vivo PK properties of IgG mutants, in which an increase or decrease in affinity to FcRn compared to WT affects column retention time.⁹ A possible drawback of FcRn-affinity HPLC combined with a measure of column retention time, however, is the lack of a direct comparison to a known in vivo parameter, which complicates the interpretation of the data within a physiological context. As FcRn binding and dissociation are largely pH-dependent, comparison of the pH of FcRn dissociation in vitro with that of antibodies in vivo is a logical physiological parameter. Although this would be possible using standard FcRn affinity HPLC, assuming a linear pH gradient across the FcRn HPLC column, measurement of the pH gradient across the column (Supplementary Figure S2a-c) shows that this is not the case. Despite not being linear, the pH gradient was shown to be reproducible and robust, with no measurable shift in the pH gradient of the mobile phase upon elution of proteins. Taken together, these observations demonstrate a rationale for coupling FcRn-HPLC to a pH monitor, to directly measure the FcRn dissociation pH. This study showed that doing so as part of FcRn-pH-HPLC allowed the measurement of the pH during column elution, and thus the pH of FcRn dissociation, facilitating a direct comparison to the optimal FcRn dissociation pH of 7.4. The FcRn dissociation pH values for cetuximab, cusatuzumab, margetuximab, durvalumab, briakinumab, adalimumab and rituximab correlated with previously established half-lives in vivo: mAbs exhibiting the most similar in vitro FcRn dissociation pH to pH 7.4 had previously shown the longest half-lives in clinical trials. This could be reflective of the mechanism of FcRn‑mediated IgG recycling. Antibodies exhibiting a relatively low FcRn dissociation pH (pH < 7.3) exhibit impaired FcRn binding compared to physiological IgG, resulting in greater trafficking to the lysosome, whereas antibodies exhibiting a relatively high FcRn dissociation pH may show impaired FcRn dissociation when exposed to pH 7.4 in the blood, leading to expedited clearance of the antibody.^10,12 Taken together, this suggests that the data obtained using FcRn-pH-HPLC are likely to have physiological relevance. However, it is important to note that FcRn-pH-HPLC, like the original assay, remains an artificial surrogate assay for potentially predicting PK properties of therapeutic antibodies. A number of key physiological factors, such as FcRn binding of serum albumin, levels of β2-microglobulin, drug-target distribution, anti-drug antibodies, which can occur as a consequence of different degrees of humanization of therapeutic antibodies, and differences in the physiological clearance mechanisms that could impact serum concentrations of IgG and FcRn binding are not considered here or incorporated into the assay. These factors may explain why cetuximab exhibits a much shorter half-life than durvalumab despite little difference in their FcRn dissociation pH.

A possible caveat of the FcRn-pH-HPLC technique compared with standard FcRn affinity HPLC is the potential for increased variability due to discrepancies in calibration of the pH monitor between independent experiments. This adds to other potential sources of variance already present using this technique, such as sample stability in the autosampler, and observed gradual degradation of active column‑bound FcRn with increasing cycles of sample injection and column regeneration. Indeed, the SD values obtained in this study do appear to be higher than those obtained in previous studies measuring relative retention time.⁹ Although a minor loss of resolution due to a small increase in variability may occur with this technique, establishing a direct comparison to physiological FcRn dissociation pH represents a substantial advantage, and any increased variance does not appear to abrogate the ability to draw a correlation between FcRn dissociation pH in vitro, and previously established in vivo half-lives of therapeutic mAbs. As such, these data suggest that FcRn-pH-HPLC could be used as a potential early surrogate assay for PK estimation in developability assessments of therapeutic antibodies.

Different modifications within a therapeutic IgG, such as the choice of target and effector antigen-binding domains, the structure in which these domains are presented, as well as the number and positioning of these domains, can all influence FcRn dissociation pH for a therapeutic IgG. The overall combination of all these factors will ultimately determine the FcRn binding properties, and thus potentially the PK profile. Given the central role of the Fc CH2-CH3 region in FcRn binding, much previous research has focused on the influence of Fc modifications on establishing differential PK properties.^17,18 A previous study aiming to identify mutations which enhance serum half-lives of IgGs found a number of Fc mutations in residues within, and peripheral to, the Fc/FcRn binding interface that influenced FcRn binding affinity, and could increase half-life 9-fold in mice and 3.5-fold in cynomolgus monkeys.⁶³ Post-translational modifications to residues within or peripheral to the IgG/FcRn binding interface have also been shown to influence FcRn affinity, with oxidation of M252 resulting in decreased FcRn affinity.^62,64 In contrast, none of the Fc mutations tested in this study, which are commonly used as Fc-silencing, Fc-enhancing, or KiH mutations in therapeutic IgGs, appeared to influence FcRn dissociation pH. However, cross-referencing the residues examined here to a previous study which characterized the degree of surface exposure of residues within the CH2 and CH3 regions of the Fc domain showed that these residues were either buried within the CH3 domain, or present within the hinge-CH2 region, and thus likely not involved in FcRn binding.⁵⁶ As such, the data presented here suggest that engineering therapeutic IgGs to hold Fc‑silencing, Fc‑enhancing, and KiH mutations can be performed without necessarily influencing the FcRn binding properties or PK profile.

This study did find, however, that altering the antigen-binding domains could significantly influence FcRn dissociation pH. Incorporating different effector domain sequences with varying affinities for CD16A significantly influenced the FcRn dissociation pH. Furthermore, FcRn-pH-HPLC analysis of bsAbs with different target domain sequences showed differences in FcRn dissociation pH when presented in Fab format. These observations are in line with previous data, which showed that altering the CDR sequences within antigen-binding domains by as much as a single residue could result in a 79-fold change in FcRn affinity, and that altered FcRn binding and PK properties are exhibited by standard IgGs with differing N-terminal target domain sequences, but identical Fc domains. ^9,23,65 Although the exact biological underpinning was not thoroughly investigated in this study, it has been suggested that altering antigen‑binding domain sequences changes the pI, and can induce conformational changes interfering with the binding between the Fc domain and FcRn affecting PK properties.^22,23,29 Indeed, IgG containing the medium affinity anti-CD16A domain, which exhibited the pI most similar to physiological pH 7.4, also exhibited the FcRn dissociation pH most similar to pH 7.4. Moreover, indication of a possible correlation between increasing pI and increased FcRn dissociation pH was also observed in identical 2Fab-Fc-2scFv bsAbs with differing target antigen-binding domains between pH 8.0 and 8.5. Although a wider pI range would need to be examined to confirm this, these data may be congruent with previous studies demonstrating a positive correlation between pI and FcRn affinity.²² Taken together, these data suggest that FcRn-pH-HPLC could potentially aid identification of optimal target and effector domains, allowing the generation of IgG-based therapeutics with optimal target affinity, whilst also balancing effective FcRn binding and a favorable PK profile.

The data presented in this study demonstrate that the number of modules, the type of modules, and the positioning of modules also significantly influence the FcRn dissociation pH. Furthermore, evidence is provided that fusion of these modules to the Fc domain of a bsAb is required for their ability to modulate FcRn binding, as the individual modules alone exhibited no or limited binding. Covalent linkage of these modules, their type and orientation, therefore, should likely be a key consideration when developing therapeutic IgGs, particularly novel bi- and multispecifics. Bispecific and multispecific antibodies are increasingly being developed as potential treatments for cancers.^59,60 As these antibodies contain multiple antigen‑binding domains, which can be incorporated as various different structures and orientations within the bi- or multispecific format, the effects of altering the type of antigen-binding domain structure used, or increasing the number of antigen-binding domains, and of altering where they occur within the molecule on FcRn binding is becoming increasingly relevant. Further study may be required to elucidate why these modifications affect the FcRn dissociation pH of a bsAb. A working hypothesis is that the most N-terminal Fab domain in the stgFab structure is further from the core FcRn binding site within the Fc domain than both of the moieties of the scDb. The scDb module is therefore closer and more likely to influence Fc/FcRn interaction. These factors should thus be considered carefully when designing novel bi- and multispecifics, allowing for balance of valency and affinity for a target or effector antigen with an optimal PK profile.

In conclusion, the FcRn-pH-HPLC method allows the interpretation of in vitro FcRn binding properties within a physiological context. Using this technique, it is possible to assess the contribution of different antibody modifications, modules, and orientations, to the FcRn retention of a candidate therapeutic IgG, thus aiding the design and developability of therapeutic IgGs with desired PK properties.

Materials and methods

Production of ROCK® engagers, half antibodies, and therapeutic reference antibodies

Generation of antibody expression vectors

Gene sequences encoding recombinant antibodies and half antibodies derived from the ROCK® platform or therapeutic reference antibodies (sequences derived from IMGT/mAb-DB)⁶⁶ were synthesized by GeneART Gene Synthesis Services (Thermo Fisher Scientific, Regensburg, Germany). Gene sequences were subcloned into the pcDNA5/FRT mammalian expression vector (Thermo Fisher, #V601020), or modified versions of the vector containing a puromycin instead of the hygromycin-resistance gene and to enable bicistronic or tetracistronic expression of product chains for multi-subunit products, using standard molecular biology techniques. In all cases, expression was driven under a human cytomegalovirus (CMV) promoter. Sequences coding for N-terminal signal peptides (MERHWIFLLLLSVTAGVHS) were added to the 5’ end of gene sequences to facilitate secretion of the antibody product. Sequences of all constructs were confirmed by Sanger sequencing at Eurofins GATC Biotech (Cologne, Germany) using custom-made primers (ELLA Biotech).

Expression and isolation of recombinant antibodies

Expression vectors encoding recombinant antibodies were co-transfected with pOG44 Flp-Recombinase Expression Vector (Thermo Fisher, #V600520) into Flp-In^TM Chinese hamster ovary (CHO) cells (Thermo Fisher, #R75807), which had been previously adapted to grow in suspension and serum-free medium (Cytiva, #SH30557.02), using Transporter 5 transfection reagent (Polysciences, #26008–5). Cells stably expressing recombinant antibodies were selected in the presence of puromycin dihydrochloride (Thermo Fisher, #A1113803) and were subsequently cultured under agitation in Erlenmeyer flasks for approximately 10 d, after which cell culture supernatants were clarified by vacuum filtration through a 0.22 µm filter and purified in a two-step procedure comprising protein A affinity chromatography and preparative size-exclusion chromatography (SEC) [half antibodies underwent protein A purification only]. For the protein A affinity chromatography, a 5 mL HiTrap MabSelect SuRe chromatography column (Cytiva, Marlborough, US, #11003495) was equilibrated in phosphate-buffered saline solution (PBS) (pH 7.4, Life Technologies, #14190–094) prior to the loading of the cell culture supernatants containing recombinant antibodies. The column was then washed with 3 column volumes (CV) PBS and 3 CV 10 mM sodium phosphate pH 7.0. Antibodies were eluted using 3 CV of 10 mM sodium acetate pH 3.5 (elution 1) or 3 CV 10 mM glycine/HCl pH 2.0 (elution 2). Elution fractions containing target antibodies were then adjusted to pH 5.0 using 1 M sodium acetate and concentrated by ultrafiltration using Amicon Ultra-15 Centrifugal Filter Units (10 kDa) (Millipore, #UFC901096) at 4°C at 3434 ×g. Preparative SEC was then performed using a HiLoad 26/600 Superdex 200 pg column (Cytiva, #28-9893-35) in SEC buffer (30 mM Sodium phosphate, 0.75 M Arginine, pH 6.0) using a flow rate of 2.5 ml/min. Elution fractions containing target antibodies were then pooled and subjected to buffer exchange using a Sephadex G-25 column (Cytiva, #17003301 or CP-0113-50 [P50] and CP‑0119‑50 [P100], Genaxxon) against 10 mM sodium acetate, 4.5% sorbitol pH 5.0. Sample purity, homogeneity, and molecular weight of purified antibodies were assessed by SDS-PAGE and analytical SEC/multi-angle-light-scattering (MALS)HPLC, and the protein content was assessed by ultraviolet-visible spectroscopy (280 nm). Purified proteins were stored at −80°C.

NIST mAb (National Institute of Standards and Technology, #8671), Cutaquig® (Octapharma, PZN 15,821,464), rituximab (InvivoSim, #SIM0008), cetuximab (InvivoSim, #SIM00021), margetuximab (Biozol #ORB746716), adalimumab (InvivoSim, #SIM0001), and briakinumab (Biozol #LT500) were purchased as indicated.

In-vitro assembly of bsAbs

Asymmetric antibodies were produced from corresponding half antibodies by in vitro assembling. Half antibodies containing a knob mutation (T366W) were mixed with corresponding half antibodies containing a hole mutation (T366S, L368A, Y407V) at equimolar concentrations. The resulting solution was titrated to pH 8.5 using 100 mM Tris-arginine pH 9.0 and supplemented with 200× molar excess of freshly prepared l-glutathione. After incubation at 32°C for 24 h, the reaction was terminated by addition of exchanging buffer to 10 mM sodium acetate, 4.5% sorbitol pH 5.0. To remove incorrectly assembled products or remaining half antibodies, target molecules were purified by SEC as described above.

FcRn‑Ph‑HPLC

FcRn chromatography was performed using a HPLC Ultimate 3000 equipped with a PCM-3000 pH and conductivity monitor (Thermo Fisher, #6082.2000). Before starting a run, a two-point calibration of the pH electrode was performed inside the flow cell with pH 4 (pH 4 ± 0.02; CarlRoth, #P712.1) and pH 9 (pH 9 ± 0.02; CarlRoth, #A519.1) buffer solution. The pre-packed 1 ml FcRn column, containing unmodified, WT human FcRn immobilized onto streptavidin-sepharose beads (Roche #08128057001) was equilibrated with five column volumes of 80% 20 mM MES, 140 mM NaCl pH 5.5 with 20% 20 mM Tris, 140 mM NaCl pH 8.8. Purified antibodies were diluted to 0.6 mg/ml in 20 mM 2-(N-morpholino) ethanesulfonic acid (MES), 140 mM NaCl pH 5.5 and 18 µg subsequently applied to the equilibrated FcRn column. The affinity-bound Fc containing molecules were eluted in 20 mM Tris, 140 mM NaCl, at increasing pH (pH 5.8–pH 8.8) in 35 column volumes. For complete elution of antibodies, the pH was held at pH 8.8 for five column volumes. All FcRn chromatography experiments were performed at 25°C at a flow rate of 0.5 ml/min. The elution profile was obtained by continuous measurement of pH and absorbance at 280 nm. The retention time was the time taken for an analyte peak to reach the detector after sample injection. The pH value at peak maximum was determined using Chromeleon software (version 7.2.10, #CHROMELEON7, Thermo Fisher). NIST mAb (National Institute of Standards and Technology, #8671), was used as control and included in every sequence. A maximum number of 30 molecules was analyzed per column (n = 3) across the column (injection scheme 30–30–30).

Abbreviations

ADCC	=	antibody-dependent cellular cytotoxicity
bsAb	=	bispecific antibody
CV	=	column volumes
DSB	=	disulfide bridges
Fab	=	fragment antigen binding
Fc	=	fragment crystallizable
FcRn	=	neonatal Fc receptor
HPLC	=	high-performance liquid chromatography
ICE	=	innate cell engager
IgG	=	immunoglobulin G
KiH	=	knobs-into-holes
NIST	=	National Institute of Standards and Technology
NK	=	natural killer
pI	=	isoelectric point
PK	=	pharmacokinetic
ROCK®	=	redirected optimized cell killing
scDb	=	single-chain diabody
scFv	=	single-chain variable fragment
stgFab	=	staggered Fabs
WT	=	wild type

Acknowledgments

The authors would like to thank Marvin Hofmann and Frederic Bethke for supporting the manuscript by conducting the experiments utilizing FcRn-pH-HPLC in the lab. The authors also thank Arndt Schottelius for providing scientific support for the manuscript.

Disclosure statement

All authors are employees of Affimed GmbH.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2023.2245519.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

The study was funded by Affimed GmbH. Medical writing assistance was provided by Meridian HealthComms, Plumley, UK, in accordance with Good Publication Practice and funded by Affimed GmbH.