Bispecific antibodies (bsAbs) are a promising class of therapeutic agents due to their ability to bind to two different types of antigens or two different epitopes on the same antigen. This ability allows bsAbs to be more effective at treating diseases. However, due to their increased complexity bsAbs often have higher levels of impurities in comparison to monoclonal antibodies (mAbs). This presents challenges to downstream processing.

We will describe how Bioprocessing Technology Institute (BTI), Singapore together with scientists at Cytiva developed an optimized capture step for two closely related knob-into-hole (KiH) bsAbs using MabSelect PrismA™ protein A resin which reached a purity of 94%. In addition, two flow-through polishing steps using orthogonal binding principles, hydrophobic interaction and hydrophobic/anion exchange, further reduced the impurities to achieve a monomeric purity of 99%.

Introduction

Because of their unique structure and increased complexity bsAbs often contain higher levels of target related byproducts and impurities, including mispaired homodimers, half-antibodies, light chain mispairings, antibody fragments, and high molecular weight species (Fig 1). These different byproducts all create challenges to downstream processing.

 

Fig 1. The two investigated bispecific antibodies and their byproducts. High-molecular weight impurities like aggregates not shown.

 

One method to reduce the amount of mispaired byproducts is to use a KiH approach, where the bsAb is designed to favor the formation of the target heterodimer (1). Hole-hole and knob-knob homodimers can still occur at low levels, with the knob-knob homodimers being rarer due to steric hindrance between the knobs. This approach reduces the levels of impurities but does not completely remove them. A purification strategy to remove the mispaired byproducts is essential. To overcome this challenge, it is possible to utilize the avidity differences inherent in an asymmetric molecule to separate mispaired antibodies, light chains, and half antibodies.

As with mAbs, protein A chromatography is one of the most commonly used affinity chromatography steps for purification of bsAbs. In particular, protein A affinity chromatography has been proposed as a method to separate bsAbs from light and heavy chain byproducts. Here we will show a three-step process used to purify two KiH bsAbs (FabscFv-KiH and Fab2scFv-KiH) with different molecular weights (Mr approximately 120 000 and 170 000 respectively) using MabSelect PrismA™ chromatography resin for capture of the respective bsAbs and two consecutive polishing steps in flow through (FT) to reach the targeted purity of monomeric purity of 99%. An intermediate pH wash step was included to remove both high molecular weight (HMW) and low molecular weight (LMW) impurities.

For more detailed information on the development of the purification process see Materials and methods

Results and discussion

Capture step

Both bsAbs showed the expected affinity to MabSelect PrismA™ resin, with a dynamic binding capacity of approximately 60 g/L resin at 10% breakthrough and a residence time of 6 min. The bsAbs were eluted at a pH of 3.6 with a decent removal of product related HMW impurities eluting in the tail of the main peak. However, as the monomeric purity in the feed material also contained significant amounts of LMW impurities co-eluting with the bsAbs, an intermediate wash step was developed. We introduced a wash step at an intermediate pH that efficiently removed the LMW impurities. The wash step was slightly different for the two bsAbs. The monomeric purity after the protein A step was above 90% with a yield close or above 90% (Table 1).

 

Table 1. Validation runs performed on 5 mL Tricorn™ 10/50 column, (6.4 cm bed height)

 bsAb stage
 Monomer concentration (mg/mL) Monomer yield (%) Momomer (%) HMW (%) LMW (%) 
FabscFv-KiH CCS* 0.69 N/A 35.5 30.8 33.8 
Wash pH 4.7
 0.02 0.6 51.5
 5.4 43.1
Elution pH 3.6 (> 50 mAU)
 5.00 91.1 93.9
 5.7 0.4 
Tail (< 50 mAU)
 0.03 1.3
 71.7
 26.0
 2.3 
Fab2scFv-KiH CCS*
 0.72 N/A 31.1
 33.5 35.4 
Wash pH 4.3
 0.01 0.4
 3.4 7.6 89.0
Elution pH 3.6 (> 50 mAU)
 2.74 89.8 92.1
 6.1 1.8
Tail (< 50 mAU) 
 0.06
 2.0 84.3
 14.2 1.5 

*cell culture supernatant

 

Polishing steps

Our hypothesis was that the complex composition of feed material would likely need at least two polishing steps after the protein A capture step. Thus, we performed a screening experiment comprising resins with orthogonal properties. The resins chosen were one hydrophobic interaction chromatography (HIC) resin (Capto™ Butyl ImpRes), one multimodal cation exchanger (MM CIEX)  resin (Capto™ MMC ImpRes), one multimodal anion exchanger (MM AIEX) resin (Capto™ adhere ImpRes), and one pure cation exchanger (CIEX) resin (Capto™ S ImpAct). We performed the screening as binding experiments using 96-well PreDictor™ plates, investigating bsAb binding capacity. To set the conditions for the study, the solubility of the bsAbs was investigated at different pH and salt concentrations. The two most promising resins were the Capto™ Butyl ImpRes HIC resin and the Capto™ adhere ImpRes MM AIEX resin. For these two resins, we identified both good binding conditions as well as non-binding conditions. We optimized the load to achieve good purity levels. The selectivity between monomer and aggregate for the chosen resins was promising.

Because the bsAbs in this study were prone to aggregate at elevated concentrations, we decided to perform the polishing steps in FT mode where concentration of bsAb is kept at low levels. The results from the 5 mL column verification experiments for the FabscFv-KiH bsAb are presented in Table 2. As the HCP levels were above 100 ppm, increasing pH above 6.5 was investigated to see if the HCP removal would improve. It was found that increasing pH to 6.8 did improve HCP removal, however, at the cost of yield.

 

Table 2. Results for the 5 mL Tricorn™ 10/50 column, (6.4 cm bed height) verification experiments at the two alternative process conditions

   FabscFv-KiH  Load
(mg/mL resin)		 Monomer concentration (mg/mL)  Step monomer recovery (%)  Overall monomer recovery (%)  HCP (ppm)   Purity (%) 
  HMW  Mono   LMW
   CSS* N/A 0.69  N/A N/A 1.36 × 106  30.8 35.5  33.8
Process 1 Post-Protein A eluate 31.5 5.00  91.1 91.1  1 779  5.7 93.9  0.4 
Polishing FT mode (Capto™ Butyl ImpRes, pH 4.0) 40 1.26 88.2  80.4 180  1.7 98.1  0.2 
Polishing (Capto™ adhere, pH 6.5) 30 0.50 83.3  66.9  135  0.5 99.3  0.2
Process 2 Post-Protein A eluate 31.5 5.00  91.1  91.1  1 779  5.7 93.9  0.4 
Polishing FT mode (Capto™ Butyl ImpRes, pH 4.0) 40 1.26  88.2 80.4  180 1.7  98.1 0.2 
Polishing (Capto™ adhere, pH 6.8) 30 0.45  75.0  60.3  81  0.4  99.4  0.2 

*cell culture supernatant

 

The results for the second bsAb (Fab2scFv-KiH) were similar to the first bsAb (FabscFv-KiH) showing that the same protocol could be used for the purification. However, some small changes had to be made to the pH of the washes to maintain the yield.

Conclusions

We have shown that:

  • MabSelect PrismA™ chromatography resin can be used to efficiently concentrate and purify two different bsAbs.
  • An intermediate wash step as well as control of loading volume/amount loaded in the flow-through steps was critical to purity for the capture step.
  • Two flow-through steps using orthogonal binding principles further reduced impurities to achieve a monomeric purity of 99%.

In addition to MabSelect PrismA™ chromatography resin, our antibody capture resin toolbox also includes MabSelect™ VL and MabSelect™ VH3 chromatography resins for purification of bsAb. The target molecule and its antibody domains determine the optimal resin. Learn more about how to select antibody capture resins.

To learn more read the full publications: Excellent removal of knob into hole bispecific antibody byproducts and impurities in a single capture chromatography (2) and Effective flow-through polishing strategies for knob-into-hole bispecific antibodies (3).

Explore MabSelect affinity capture resins and Capto polishing resins  for purification of bispecific antibodies.

  1. Ridgway JP, Presta LF, Carter P, ‘Knobs-into-holes’ engineering of antibody CH3 domains for heavy chain heterdimerization. Protein Eng. 1996;9(7):617-621.
  2. Chen SW, Hoi KM, Mahfut FB, Yang Y, Zhang W, Excellent removal of knob into hole bispecific antibody byproducts and impurities in a single capture chromatography. Bioresources and Bioprocessing. 2022;9(72).
  3. Chen SW, Hoi KM, Mahfut FB, Yang Y, Zhang W, Effective flow-through polishing strategies for knob-into-hole bispecific antibodies. Bioresources and Bioprocessing. 2022;9(98).

bsAb culture

CHO K1 cell lines producing FabscFv-KiH and Fab2scFv-KiH were expressed and harvested at an approximate titer of 0.5 g/L. The bsAbs were engineered to form KiH to facilitate heterodimeric Fc pairing, minimizing the impurity levels.

Optimization of the capture step

The following workflow was used to optimize the capture step using MabSelect PrismA™ resin:

  1. Determination of the binding capacity
  2. Determination of the load and residence time
  3. Optimization of elution pH
  4. Optimization of the wash conditions

 

Determining the MabSelect PrismA™ resin binding capacity for the bsAbs

The experiments were performed using a 1 mL Tricorn™ (5/50) chromatography column with a bed height of approximately 5 cm to minimize the usage of the material. Neutral pH binding conditions, that is, PBS-like conditions, were used. The flowthrough (FT) was collected, and each fraction was analyzed by SEC-HPLC and the feed background was subtracted from the signal. The breakthrough curve is shown in Figure 2. The dynamic binding capacity (DBC) was determined at 10% breakthrough (QB10%) to be 63 mg/mL for FabscFv-KiH and 61 mg/mL for Fab2scFv-KiH.

 

Fig 2. Breakthrough curves for the two bsAbs at 6 min residence time, PBS-like conditions, and MabSelect PrismA™ resin. Concentrations were determined using SEC subtracting the feed background from the signal.

 

Optimizing load conditions and residence time for purification of bsAbs using MabSelect PrismA™ resin

Several runs were performed to optimize the load and residence time (RT) (Table 3). The RT was set to 6 min to enable scale up to a larger column. The load was set to 50% QB10% as it was found that the bsAbs aggregated at higher loads.

 

Table 3. Effect of different loads and RT on purity and recovery of the two bsAbs eluted at pH 3.6 using MabSelect PrismA™ resin

 

bsAB

 

Sample

 

Loading conditions

 

 

Monomer concentration
(mg/mL) 

 

Monomer recovery
(%)

 

Purity
(%) 
     QB10 (%)  RT (min) HMW  Monomer LMW
 
 
FabscFv-KiH		 CSS* N/A N/A  0.52  N/A  29.5  28.3 42.2 
MabSelect
PrismA™
resin eluate		 80 6  2.77 82.4 14.2  85.3  0.5 
50  6 2.59 90.4 11.0  88.5 0.5
50 2 2.59 90.4 11.2  88.3  0.5
 Fab2scFv-KiH CSS* N/A N/A 0.72 N/A  31.1 33.5 35.4 
MabSelect
PrismA™
resin eluate		 50
 6
 2.14 91.2 8.4 89.1 2.5 

*cell culture supernatant

 

Optimizing elution pH for purification of bsAbs using MabSelect PrismA™ resin

The optimal pH for the step elution was evaluated using a gradient elution from pH 6.0 to 3.0 (50 mM sodium citrate) over 25 column volumes (CVs) at 2 min RT (Fig 3). Fractions from the elution was collected and analyzed by SEC (Fig 3) and SDS Page (not shown). As illustrated in the SEC analysis there are impurities present in the main peak, in addition to the step elution, which indicates that the purity might increase if a wash step is performed before the elution.

 

 

Fig 3. A) Elution peak for purification of FabscFv-KiH bsAb using MabSelect PrismA™ chromatography resin. B) SEC analysis of fractionated peak in A. The mispaired hole-hole Ab (approximately Mr 106 000) is marked with an asterisk (*) and elutes right after the correct bsAb (approximately Mr 125 000) in SEC and is most abundant in the first fraction, that is, before the main peak in the elution. The large peak eluting at 6 mL from the SEC column is most abundant in the late fractions from MabSelect PrismA™ resin and its position in the analysis indicates that this is aggregated forms of the bsAb. The peak at approximately 7 min corresponds to the knob-knob Ab (Mr 146 000).

 

Optimization of wash conditions

We investigated the potential of a wash step to remove some of the impurities, such as half antibodies and smaller hole-hole homodimers, as they eluted early in the pH gradient. After optimization of the wash step, we were able to improve the purity obtained from the capture step from 88% to more than 93% with a monomeric yield above 90%.

For the second bsAb (Fab2scFv-KiH) similar experiments were performed and almost the same conditions were found. The pH in wash was lowered to 4.3, compared to 4.6 for the FabscFv-KiH bsAb, to get good removal of impurities. Also, the Fab2scFv-KiH bsAb eluted at pH 3.6 with a monomeric yield of approximately 90%, and a bsAb purity level above 90%.

 

Development of the polishing step

The following steps were used to optimize the polishing step:

  1. Stability study at tentative binding and elution conditions
  2. Screening of resins using 96-well PreDictor™ plates
  3. Investigation of conditions for bind/elute (B/E) and FT
  4. Initial column experiments
  5. Column experiments at tentative FT conditions

 

Stability study of bsAb at tentative binding and elution conditions

The bsAbs stability was investigated at a concentration of 5 mg/mL in a pH interval of 4 to 8 and in a NaCl concentration from 0 to 500 mM.  It was found that the full pH-range could be used but that aggregation occurred at salt concentrations above 400 mM.

Screening of resins using 96-well PreDictor™ plates

We selected to investigate four different resins for the polishing step. One strong cation exchanger (CIEX) (Capto™ S ImpAct), one multimodal resin with cation exchange properties (MM CIEX) (Capto™ MMC ImpRes), one multimodal resin with anion exchange properties (MM AIEX) (Capto™ adhere ImpRes for the plate study and Capto™ adhere for the column study), and one hydrophobic interaction chromatography (HIC) resin (Capto™ Butyl ImpRes). Capto™ adhere and Capto™ Butyl ImpRes chromatography resins were the most promising resins. Binding capacity was determined in over-load mode for both resins, as they might be operated in FT mode.

Investigation of conditions for B/E and/or FT Capto™ Butyl ImpRes

Figure 4 shows the static binding capacity (SBC) for the FabscFv-KiH bsAb on Capto™ Butyl ImpRes chromatography resin where citric buffer of different concentrations has been used to control conductivity and pH. From this plot we concluded that for binding the bsAb, the salt concentrations needed would probably cause aggregation of the bsAb. Therefore, this step should be operated in FT mode, that is, concentration of sodium citrate, should be kept low or intermediate in the range studied. When looking at the % of HMW aggregates in the FT from the binding experiment, we found that the amount of aggregates in FT depends on salt and pH (Fig 5). Compiling the information from Figures 4 and 5 results suggested pH 4.0 at 50 mM sodium citrate as the conditions for the FT step.

 

Fig 4. SBC for the FabscFv-KiH bsAb on Capto™ Butyl ImpRes resin using a 96-well 6 µL PreDictor™ plate loaded at 2 g/L (200 µL). Salt concentration in this plot means concentration of sodium citrate.

 

 

Fig 5. Presence of HMW aggregates (% HMW) in FT for the loading phase of Capto™ Butyl ImpRes resin. The 96-well PreDictor™ plate was loaded at 2 g/L (bsAb volume 200 µL, ~ 10% aggregate). The blue area corresponds to low content of aggregates in FT.

 

Initial column experiments using Capto™ Butyl ImpRes chromatography resin

From the PreDictor™ plate studies, low pH and buffer concentration was found to provide promising conditions for the removal of aggregates in FT mode, and to achieve a high yield of bsAb. This knowledge was used to run the column experiment, where bsAb was loaded at pH 4.0 with 50 mM sodium citrate. The experiment was performed in Tricorn™ 5/50 1 mL column (5 cm bed height). The results (monomer yield, HMW content, and LMW content) of continuous loading of bsAb up to 140 mg/mL resin while collecting fractions are shown in Figure 6. Some pools were also analyzed for HCP and as can be seen from Table 4 there was breakthrough of HCPs at elevated loads and thus it was decided to use a load of 40 mg/mL resin.

 

Fig 6. Result of continuous loading FT experiment performed on a Tricorn™ 5/50 1 mL column with Capto™ Butyl ImpRes chromatography resin. Fractions were analyzed with SEC and results is presented as monomer (bsAb) and HMW and LMW, all as cumulative amounts or relative amounts. The load contained about 93% monomer, 6% HMW, and 0.5% LMW.

Table 4. Result of continuous loading experiment for purification of bsAb on Tricorn™ 5/50 1 mL column with Capto™ Butyl ImpRes chromatography resin

Load (mg/mL resin)

Monomer concentration
(mg/mL)

Monomer yield 
(%)

HCP (%)

Purity

(%) 
  HMW  Mono LMW
20 
 0.93 81.1  235  0.8  98.3  0.9 
40 1.34  90.8  243 1.3 98.0  0.7 
75 1.54 94.4 1114  2.1 97.4 0.6 
117  1.78  97.1  1713  2.7 96.4 0.9 

 

Investigation of conditions for B/E and FT

Capto™ adhere ImpRes was identified as a potential resin for the second polishing step. The investigated pH range was 5 to 6.5 with the intention to also run this resin in FT mode. 96-well PreDictor™ plate experiments were performed with loading a 200 mL sample in concentrations of 2.5 mg/mL and a resin volume of 6 µL per well. The experiments were conducted with a sample that was pre-purified on MabSelect PrismA™ resin with an aggregate content of approximately 12%.

The PreDictor™ plate experiments were used to choose the conditions for the column experiments with the process run in FT mode, that is, the intention was to find conditions where the aggregates bind to the resin while the desired bsAb pass through the column. The highest binding conditions for aggregates occurred when pH was in the higher range, that is, pH 6 to 6.5, and the concentration of salt was low (Fig 7). 
The high pH and low salt concentration which was good for binding HMW aggregates was also where the bsAb bound strongly to the resin. Although it seems like the aggregates bind stronger at lower pH, FT conditions could not easily be chosen. Therefore, to optimize the process conditions, initial column experiments were also performed.

 

Fig 7. Result of 96-well PreDictor™ plate experiments for SBC for HMW at a monomer (A) load and for bsAb (B) for Capto™ adhere ImpRes chromatography resin at sample concentrations of 2.5 mg/mL.

 

 

Fig 8. Results for gradient elution of bsAb on Capto™ adhere resin peak max is obtained at pH 5.6.

 

Table 5. Result of fraction analysis of B/E experiment from Capto™ adhere chromatography resin

Fraction (0.5 mL)

pH in fraction

bsAb in fraction (mg)

bsAb purity in fraction (%)

HMW in fraction (%)
1B8 6.48 0.013  100 0
1B9 6.43 0.021 100  0 
1B10 6.37  0.033 100 0 
1B11 6.31 0.055  100 0 
1B12 6.21 0.09 99.8  0 
1C1 6.17 0.136  99.4 0.6
1C2 6.11 0.18 99.6 0.4
1C3 6.03  0.222 99.5 0.5
1C4 5.96 0.254 99.4  0.6
1C5 5.89 0.271 99.4 0.6
1C6 5.81 0.283 99.4 0.6
1C7 5.73 0.284  99.4 0.6
1C8 5.67  0.284  99.4 0.6
 1C9 5.59 0.276 99.4 0.6
1C10 5.51 0.268  99.4 0.6 
1C11 5.43 0.262  99.4  0.6
1C12 5.36 0.26  99.4 0.6
1D1 5.28 0.254  99.1  0.9 
1D2 5.17  0.229 96.8 2
1D3 5.07 0.21 94.4 4.4
1D4 4.97 0.181 91.7  6.9
1D5 4.87 0.14 90.2 8.4
1D6 4.76  0.102 89.9 8.6
1D7 4.66  0.074 89.9 8.7
1D8 4.55 0.055 89.3  9.1 
1D9 4.45  0.043 91.2 7.2 
1D10 4.35 0.033 91.8 6.8 
1D11 4.25  0.025  93.7 5.2 
1D12 4.15 0.019 94.4  4.7 

 

Plotting the cumulative yield of monomer versus yield of HMW aggregates shows that the HMW aggregates bind strongly to the resin and that the original content of > 10% HMW aggregates was reduced down to 2% when collecting the full peak (Fig 9).

 

 

Fig 9. Low load experiment of 5 mg/mL Capto™ adhere chromatography resin presented as cumulative HMW versus cumulative yield of monomer.

 

Column experiments at tentative FT-conditions

Following the initial column experiments, additional column experiments were performed at conditions close to the elution pH, that is, at pH 5.7 (Fig 10), and for comparison also at a pH 6.5 where the HMW aggregates did not elute according to the initial column experiment. In both cases the load was 30 mg/mL resin and ran in FT mode and fraction ware collected and analyzed for target molecule and impurities. The start material concentration varied for the two pHs (both close to 1 mg/mL), but the composition was very similar. As can be seen in Table 6 the highest yield was obtained at the lower pH. The impurity profile was similar at both conditions, but the lower pH resulted in more LMW and HMW aggregates.

 

 

Fig 10. Chromatogram from a load of 30 mg bsAb/mL resin. The FT was collected as well as the wash fraction.

 

The 1 mL column experiments were repeated, and analysis of the HCP content was performed (Table 6). As can be seen in the table, the HCP removal was less efficient using the lower pH however the HCP levels have to be balanced with yield.

 

Table 6. 1 mL column experiment results using Capto™ Butyl ImpRes and comparing the use of two different pH-conditions on Capto™ adhere

 

FabscFvKiH Capto™ Butyl ImpRes FT

FabscFvKiH Capto™ adhere FT
pH 5.7 Peak + 25 CV wash

FabscFvKiH Capto™ adhere FT
pH 6.5 Peak + 25 CV wash
FabscFv-KiH concentration (mg/mL)  1.34  0.53  0.38 
FabscFv-KiH yield  (%) N/A  95.1  79.1 
Purity (%)  98.0 98.4 98.9 
HMW (%)  1.3 0.7  0.5 
LMW (%)  0.7 1.0 0.7 
HMW (ppm) 187 130  99