HTS and Process Development for Capture of Recombinant Pro-insulin from E. coli using Capto MMC

Extracted from Multimodal Chromatography (PDF), GE Healthcare, 2013

This study describes the development of a robust capture method for recombinant pro-insulin.

PreDictor plates and Assist software were used to determine a chromatographic medium for capture and identify promising binding and elution conditions. Based on the screening results, Capto MMC was selected as the most promising medium due to its ability to bind sample without prior dilution. With the binding and elution conditions found in the screening experiments as the starting point, the capture step was optimized on a Tricorn 5/50 column packed with 1 mL of medium. Once a robust capture protocol had been established, the process was successfully scaled up from Tricorn to HiScreen prepacked columns, HiScale 16/40 column (20 cm bed height) packed with Capto MMC, and finally to an AxiChrom 50/300 column (19.5 cm bed height, 400 mL packed bed volume).

Materials and methods

Screening with PreDictor plates

Whenever possible, experiments with PreDictor plates were performed with fully automated protocols on a Tecan Freedom EVO-2 200 Robotic System. More complex protocols such as sample handling were carried out manually. Liquid removal was performed by vacuum or centrifugation throughout the study.

The pro-insulin used in all experiments originated from E. coli. It was supplied by BIOMM S.A., Belo Horizonte, Brazil. The pro-insulin solution was subjected to sulfitolysis to hinder the formation of disulfide bridges. The suspension that contained 8 M urea was approximately 10 mg/mL in recombinant pro-insulin and 18 mg/mL in total protein. The conductivity of the sample was approximately 14 mS/cm. PreDictor experiments followed the illustration shown in Figure 4.39. Conditions studied are presented in “Results and discussion.”

Schematic illustration of the workflow of a batch experiment in the wells of a PreDictor plate.

Fig 4.39. Schematic illustration of the workflow of a batch experiment in the wells of a PreDictor plate. The same steps would be employed in a column experiment, that is, equilibration, sample addition, wash, and elution. The gray color in the wells represents the chromatography medium; red shades (red and pink) represent different concentrations of protein solution. Brown represents the medium with bound sample.

Analysis

In the PreDictor binding studies, capacities were measured from analyses of the flowthrough fraction. In the elution studies, the first elution fraction was evaluated. Start samples were analyzed in all studies. All analyses were performed by AIEX on a Mono Q™ 5/50 GL column. The pro-insulin sample concentration was determined by integrating the area of the peak eluting at a retention time of 9 to 10 min and relating its surface area to that in the crude sample:

 

The resulting pro-insulin concentration in the flowthrough or first elution fraction for each condition was used as in-data in Assist software where the response surfaces for experimental evaluation were generated.

Column experiments

Column experiments comprising optimization, DBC experiments, a robustness study, and scale-up, were performed with the Capto MMC multimodal medium on chromatography systems suitable for the column dimensions. Table 4.12 summarizes the columns, systems, samples, and purification conditions for these experiments.

All eluent buffers were prepared in 8 M urea and all experiments were concluded with 1 M NaOH CIP followed by storage in 20% ethanol. Detection was performed at 280, 405, and 260 nm. In the preliminary elution experiments, salt/pH gradients were used while optimization, robustness, and scale-up experiments were performed as step elutions. As in the PreDictor experiments, sample analyses were performed by the previously described Mono Q method.

Table 4.12. Summary of the Capto MMC column experiments in process development

Study Column Vc (mL) Sample load (mL) System Flow rate (mL/min)
DBC Tricorn 5/50 1 10 ÄKTAmicro™ 0.2
Elution optimization Tricorn 5/50 1 2.5 ÄKTA avant 25 0.2
Robustness study Tricorn 5/50 1 2.5 ÄKTA avant 25 0.2
Scale-up 2 × HiScreen 4.7/10 9.4 24 ÄKTA avant 25 1.9
  HiScale 16/40 40 100 ÄKTA avant 150 8
  AxiChrom 50/300 400 960 ÄKTA avant 150 80

Results and discussion

Screening experiments for binding

Binding experiments were performed on a selection of ion exchange and multimodal media; the PreDictor plates contained 2 μL or 6 μL of SP Sepharose Fast Flow, Capto S, or Capto MMC. The small media volumes (2 and 6 μL) enabled binding experiments by overloading the media with the buffered sample (200 μL solution 2.5 mg/mL in respect to pro-insulin per well) in 8 M urea without consuming more than 15 mL of crude sample for the binding study. The binding with respect to both the initial salt concentration and the pH value of the binding buffer was examined. Table 4.13 summarizes the test conditions for each medium.

Table 4.13. Summary of media and parameters in the binding experiments conducted on PreDictor plates

Experiment pH NaCl (mM)
Binding study—CIEX screening plate
-Capto S, 2 μL
-Capto MMC, 6 μL
-SP Sepharose Fast Flow, 6 μL
3.4–5.0 0–300
Binding study—Capto MMC, 6 μL 3.0–7.0 0–300

In all binding experiments, the flowthrough fraction was collected and analyzed with respect to nonbound pro-insulin as compared with the start sample, which gives an indication of the binding capacity at each condition. The resulting response surfaces for all media, generated using Assist software, are shown in Figure 4.40.

Capto S and SP Sepharose Fast Flow indicate high binding capacities at the lowest pH tested (i.e., 3.4) and no salt. Capto MMC binds at 150 mM salt and higher pH compared with these two media. Because the starting sample of the fusion protein has an ionic strength close to 150 mM NaCl, high binding capacity at this concentration is an advantage.

A second binding study was thus performed with Capto MMC and a broader parameter interval intended to reveal the optimum binding for this medium. As Figure 4.41 shows, the highest binding capacities (red/orange zone) for pro-insulin binding to Capto MMC are obtained at pH 5 (or just above) and 0 to 160 mM NaCl. It was decided to continue with Capto MMC and to study conditions for elution.

Fig 4.40. Response surfaces generated by Assist software for pro-insulin binding (g/L) as a function of NaCl concentration (x-axis) and buffer pH (y-axis) for SP Sepharose Fast Flow, Capto S, and Capto MMC, respectively.

Fig 4.40. Response surfaces generated by Assist software for pro-insulin binding (g/L) as a function of NaCl concentration (x-axis) and buffer pH (y-axis) for SP Sepharose Fast Flow, Capto S, and Capto MMC, respectively. The range of binding capacities achieved is shown to the right of each surface. Black crosses represent actual data between results that have been interpolated.

Fig 4.41. Response surface for binding (g/L) of pro-insulin on Capto MMC as a function of NaCl concentration (0 to 300 mM) and buffer pH (4 to 7.5). Assist software was used in visualizing this data.

Fig 4.41. Response surface for binding (g/L) of pro-insulin on Capto MMC as a function of NaCl concentration (0 to 300 mM) and buffer pH (4 to 7.5). Assist software was used in visualizing this data.

Screening experiments for elution

PreDictor plates with 50 μL of Capto MMC media volume were used for elution studies. This ensured sufficient loading to detect the target molecule without overloading the medium. The amount of protein applied in the loading step corresponded to 70% of the binding capacity that was estimated in the binding study, that is, 180 μL of sample, 5 mg/mL in respect to pro-insulin. The elution study was performed using a range of eluent compositions: pH 3.7 to 7.6 and 150 to 1000 mM NaCl. The evaluation procedure was the same as for the binding study, but now the first elution fraction was analyzed. This showed the conditions required to obtain elution in the column verification work that followed.

As only the first elution fraction was analyzed, one may not expect full yield in this step. The highest yield achieved was 70% and was found at pH 7.5 and a NaCl concentration above 600 mM (Fig 4.42).

Fig 4.42. Pro-insulin yield (%) (first elution fraction) on Capto MMC as a function of NaCl concentration (150 to 1000 mM) and buffer pH (4 to 7.5). Assist software was used in obtaining these data.

Fig 4.42. Pro-insulin yield (%) (first elution fraction) on Capto MMC as a function of NaCl concentration (150 to 1000 mM) and buffer pH (4 to 7.5). Assist software was used in obtaining these data.

Optimization in Tricorn columns on ÄKTA avant 25

The HTS experiments on PreDictor plates suggested that the best conditions for pro-insulin capture would be binding at around pH 5 and a NaCl concentration of 50 to 150 mM on Capto MMC followed by eluting at a pH greater than 7 and a NaCl concentration above 600 mM.

With these parameters added as factors in a DoE protocol, the capture step was optimized on Capto MMC packed in a 1 mL Tricorn column (diameter 5 mm). As 150 mM NaCl corresponds to the isotonic salt concentration found in the start sample, this salt concentration was an obvious starting point for binding because it eliminated the need to dilute sample prior to loading. Binding buffer pH was set at 5.2, and the pH of the start sample was set accordingly. In the first column experiment, elution with a salt gradient was tested by loading 20 mg of proinsulin (2 mL sample) in 50 mM sodium acetate buffer, pH 5.2 in 8 M urea on the 1 mL column and eluting with a linear salt gradient of 150 to 1000 mM NaCl for 7 CV.

Figure 4.43 shows the results. The fraction collected at the maximum height of the elution peak was analyzed on Mono Q and the resulting chromatogram compared with that of the crude sample and one flowthrough fraction.

Analysis of the flowthrough fraction (Fig 4.43B) showed good binding of the target molecule with no pro-insulin detected in the flowthrough. The nonprotein impurity seemed to be low binding as it appeared in the flowthrough fraction while the corresponding peak in the elution fraction was significantly smaller. This indicated good capture and purification of pro-insulin.

However, a large peak was seen during CIP (Fig 4.43A), suggesting that high salt concentration alone was not adequate to recover all of the pro-insulin. Experience with several other target proteins indicates that multimodal media frequently require more than just high ionic strength for efficient elution.

alttext

Fig 4.43. (A) A 2 mL crude sample, pH 5.2 in 8 M urea, loaded on a Tricorn 1 mL 5/50 column packed with Capto MMC and eluted by a linear salt gradient from 150 to 1000 mM NaCl for 7 CV. (B) Corresponding Mono Q analysis of crude sample, flowthrough, and one elution fraction (collected at the main elution peak maximum). In both A) and B), detection was at 280 nm.

Figure 4.44 shows the capture and analysis results where the salt gradient was supplemented with a pH 5.2 to 7.5 gradient.

Fig 4.44. (A) A 2 mL crude sample, pH 5.2 in 8 M urea, loaded on a Tricorn 1 mL column packed with Capto MMC and eluted with a linear combined salt and pH gradient from 150 to 1000 mM NaCl and pH 5.2 to 7.5 for 7 CV. (B) Corresponding Mono Q analysis of the flowthrough, wash, and pooled fractions in the main elution peak. In both A) and B), detection was at 280 nm.

Fig 4.44. (A) A 2 mL crude sample, pH 5.2 in 8 M urea, loaded on a Tricorn 1 mL column packed with Capto MMC and eluted with a linear combined salt and pH gradient from 150 to 1000 mM NaCl and pH 5.2 to 7.5 for 7 CV. (B) Corresponding Mono Q analysis of the flowthrough, wash, and pooled fractions in the main elution peak. In both A) and B), detection was at 280 nm.

Comparing chromatograms for the constant pH (Fig 4.43A) and the pH gradient (Fig 4.44A) capture experiments revealed that a combined pH and salt gradient gave both a narrow elution peak and a high yield, neither of which was achieved when salt gradient elution alone was employed.

DBC experiments

Once promising conditions for binding and eluting pro-insulin had been established, attention was turned to DBC. This was determined by overloading the column with crude sample and collecting and analyzing fractions to determine the point at which pro-insulin breakthrough occurred. Based on DBC experiments, the loading in the experimental work was set to 25 mg pro-insulin (2.5 mL crude sample, approx. 80% of DBC) to secure complete binding.

Elution optimization

Aiming at a step elution mode, elution conditions were optimized using the buffer prep and DoE tools of ÄKTA avant 25. A full factorial design with three center points based on two variables (pH and NaCl concentration) each at three levels was set up to determine the salt concentration and the pH needed to obtain sufficient purity and yield (above 80% and 95%, respectively). The area of the pro-insulin peak as well as the area percent of pro-insulin in the analysis chromatogram (purity) were set as responses. See Table 4.14 for details.

Table 4.14. Design variables, values for the elution optimization, and purity data of pro-insulin in the eluted peak

Run NaCl (mM) Elution pH read2 Area (mAU × mL) Purity (%)
11 450 7.1 196 76
2 150 7.1 181 79
31 450 7.1 196 81
4 150 8 246 83
51 450 7.1 187 82
6 750 8 251 84
7 750 7.1 241 82
8 450 6.2 69 71
9 450 8 250 84
10 150 6.2 37 53
11 750 6.2 116 78

1 Center points.

2 8 M urea influences the pH reading; settings in ÄKTA avant were approximately 1 pH unit lower.

Figure 4.45 shows the pro-insulin peak area in the collected elution peak as a function of pH and NaCl concentration. This clearly demonstrates that the optimal elution for pro-insulin is found at high pH, whereas an increase in the concentration of NaCl above 150 mM has only a minor effect. The purities achieved were also highest at high pH. It was decided to perform the elution at pH 8 and 150 mM NaCl.

Fig 4.45. Response surface for the elution peak area of pro-insulin as a function of pH and NaCl concentration in mM. R2 (explained variation) = 0.989, Q2 (predicted variation) = 0.736. ÄKTA avant was used in obtaining these data.

Fig 4.45. Response surface for the elution peak area of pro-insulin as a function of pH and NaCl concentration in mM. R2 (explained variation) = 0.989, Q2 (predicted variation) = 0.736. ÄKTA avant was used in obtaining these data.

Robustness study

To conclude process development, a robustness study was performed on 1 mL Tricorn 5/50 columns packed with Capto MMC using the optimized elution conditions of pH 8 and 150 mM NaCl. The robustness study was designed using a Plackett–Burman DoE based on four variables (two chromatography media batches, two crude sample batches, elution pH 7.8 to 8.2 and load volume 2.3 to 2.7 mL) with 150 mM NaCl in all eluent buffers. Figure 4.46 shows the scaled and centered coefficients for the purity data (all above 80% purity) from the eluted peaks as a function of the variable parameters. It is clear that no significant model terms can be detected. The yield was approximately 95% for all conditions in this study.

alttext

Fig 4.46. Scaled and centered coefficients for purity as a function of four variable parameters: chromatographic medium lot (×2), sample (×2), pH, and load volume.

Scale-up experiments

Columns with 20 cm bed heights were used for 9-, 40-, and 400-fold scale-up by increasing column diameter while keeping other parameters such as residence time and sample load/mL chromatographic medium constant. Other conditions were similar to those found in the optimization study on the Tricorn 5/50 column packed with Capto MMC (loading at pH 5.2 and elution at pH 8, both in the presence of 150 mM NaCl).

Two HiScreen Capto MMC columns were connected in series to give 20 cm bed height. In addition, a HiScale 16/40 column (diameter 16 mm) and an AxiChrom 50/300 column (diameter 50 mm) were packed with Capto MMC to bed heights of 20 and 19.5 cm, respectively. The capture experiment was performed at 240 cm/h at all three extended scales (5 min residence time). Fractions from the flowthrough and the eluted peaks were analyzed on the Mono Q column. Results and purity data (Table 4.15 and Fig 4.47) show that the capture step of the pro-insulin purification was successfully transferred from the 1 mL Tricorn 5/50 column to the 400 mL AxiChrom 50 column. The resulting pro-insulin purity was 82%, and the yield was 96% measured at the 400 mL scale.

Table 4.15. Purity data after four column steps

Column1 Scale-up factor Crude sample load (mL) Pro-insulin purity (%)
Tricorn 5/50 1 2.5 83
HiScreen Capto MMC × 2 9.4 23.5 86
HiScale 16/40 40 100 84
AxiChrom 50/300 400 960 82

1 Total packed bed heights were 20 cm, except for AxiChrom 50, which was 19.5 cm.

 

Fig 4.47. Chromatogram from the final (AxiChrom column) step after preceding columns.

 

Fig 4.47. Chromatogram from the final (AxiChrom column) step after preceding columns. The pro-insulin crude sample was loaded on Capto MMC at pH 5.2 and 150 mM NaCl in a HiScreen (9 mL), HiScale (40 mL), and AxiChrom 50/300 (bed height 19.5 cm; 400 mL) column. An ÄKTA avant 150 system was used with the AxiChrom column.

 

Conclusions

HTS with PreDictor plates and the Assist software allowed quick selection of most suitable chromatography medium and identification of promising binding and elution conditions for the capture of recombinant pro-insulin expressed in E. coli. This gave a fast and confident start to the purification process development.

Based on these screening experiments, Capto MMC was the medium of choice for further work due to its ability to bind sample without prior dilution. The capture step was further optimized in a Tricorn 1 mL column packed with Capto MMC, again based on the binding and elution conditions determined by the screening experiments. Once the optimized protocol had been confirmed to be robust, the process was successfully scaled up from a 1 mL Tricorn 5/50 column to a 400 mL AxiChrom 50/300 column. The resulting purity for the capture step was 82% with a yield of 96%.

The overall outcome demonstrates the value of introducing high-throughput methods into process development workflows. In this PreDictor plate screening example, media and condition selection was completed in 1 wk using 30 mL of crude sample (300 mg of the target molecule). When UV absorbance in a plate reader is sufficient for evaluation, media screening can be finalized within two days. The screening described here enabled fast development of a pilot-scale process (400 mL AxiChrom column) within 4 wk.

For more information on this example, see application note 28-9966-22, “High-throughput screening and process development for capture of recombinant pro-insulin from E. coli.

Materials

     
Related Links