Overcoming Challenges in the Reconstitution of a High-Concentration Protein Drug Product
A version of this article previously appeared in the March 2013 issue of BioPharm International.
The protein in the current study discussed here is a proprietary pharmaceutical recombinant protein with a molecular weight of approximately 210 kDa. Consisting of 40 mg/mL protein and certain common excipients (Table I), the Formulated drug substance (FDS) was lyophilized as the solid drug product (DP). For subcutaneous (SC) administration of the 80 mg/mL protein dosage form, end users are required to reconstitute the DP with water for injection (WFI) to half the original fill volume. With the recommended reconstitution method of continuous swirling, complete reconstitution of the DP can be variable and lengthy between 4 - 15 min.
During the Phase II clinical trial for rheumatoid arthritis, patients complained that the reconstitution of the DP was too time-consuming and hence painful. Consequently, the lengthy reconstitution needed to be shortened to facilitate patient compliance and ease-of-use. Because the contract manufacturing for the Phase III clinical trial DP would start in 4 months, any modifications to the formulation and/or process had to be implemented within that time frame. If the problem remained by then, a mechanical orbital shaker would be supplied to each patient to replace the reconstitution by hand.
The work to reduce the reconstitution time reported here may provide useful information and practical guidelines for the development of other high-concentration protein drug products.
Table I. The Formulation
|Protein||40 mg/mL||80 mg/mL|
|Buffer||20 mM||40 mM|
|Amino acid 1||0.5%||1.0%|
|Amino acid 2||25 mM||50 mM|
|Final volume||5.5 mL||2.75 mL|
Materials and Methods
FDS at a protein concentration of 40 mg/mL was filtered through a 0.22-µm filter (Millipore) and 5.5 mL was filled in 20-mL vial (West Company). Chamber pressure held at 100 mTorr was controlled with a capacitance manometer. Chamber moisture was monitored with a Pirani gauge. Product temperature was monitored with thermocouples placed at the bottom center of the vial. Lyophilization was carried out in a LyoStar II freeze-dryer (FTS Systems). Unless otherwise specified, the cycle consisted of cooling at 5°C for 60 min and freezing by lowering the temperature by 0.5°C/min to -45°C for 90 min. Primary drying was initiated by first evacuating to 100 mTorr followed by raising the shelf temperature by 0.5°C/min to -5°C. The primary drying was continued until the Pirani pressure deceased to 100 mTorr. For the secondary drying, the shelf temperature was increased by 0.7°C/min to 25°C for 6 h. Vials were stoppered under a 608,000 mTorr anhydrous nitrogen gas pressure.
For reconstitution, 2.3 mL WFI was injected into the vial with a 3 mL BD syringe. Unless otherwise stated, the dissolution was observed visually as the vial was swirled by hand.
FDS at a protein concentration of 40 mg/mL was diluted 1.6-fold with WFI, filtered as above, and 8.8 ml was filled in 20-mL vial, stoppered, and weighed. Shelf temperature ramp rate was constant at 0.5°C/min in all steps below.
There were 5 vials per group. Each group was separately cooled at 5°C for 1h and frozen at -40°C for 2 h prior to initiating the annealing described below. The control group without annealing was transferred to a -40°C freezer at this point.
(A) To determine the effects of annealing temperature on the primary drying rate, each of the four sample groups was annealed for 3 h at either -3°C, -8°C, -13°C or -18°C. Samples were re-solidified at -40°C for 1h and transferred to the -40°C freezer.
(A2) At the completion of the annealing for all the groups, the vials including the controls were transferred from the -40°C freezer to the -40°C shelf in the lyophilizer, surrounded by two layers of placebo-containing vials, and held at -40°C for 1h. Primary drying was initiated by first evacuating to 100 mTorr and then raising the shelf temperature to -5°C. Vials were stoppered after 3.5-4.5 h when 20%-40% of the crystalline water had sublimed. Samples were re-weighed, and the primary drying rate was calculated using the weight loss during the partial drying.
(B) To determine the effects of annealing time on the primary drying rate, each of the three sample groups was annealed at -8°C for 1h, 2.5 h or 4 h. Samples were re-solidified at -40°C for 1 h and transferred to the -40°C freezer. Then the lyophilization procedure above was followed.
SE-HPLC was performed using an Agilent 1100 system and detection was at 215 nm. A TSK-Gel G4000SWXL column (TOSOH Bioscience) was used with mobile phase consisting of 0.5N NaCl, 10 mM sodium phosphate, pH 7.4, at a flow rate of 0.4 mL/min. Mass load of the protein was 25 µg. In preliminary experiments with multi-angle light scattering and refractive index detectors connected in series, the molecular weight (MW) of eluted species was determined. The main peak eluted at ~21 min was assigned to native monomeric protein. Higher MW species eluted at ~19 min were defined as soluble aggregates, and lower MW species eluted at ~23 min as degradation products. Empower 2 software (Waters) quantifies each as % of the total protein.
Samples were analyzed on an ABB FTLA 2000 spectrometer equipped with a DuraSampleIR II ATR. Prota (Grams/32) software was used for data acquisition, subtraction of buffer & vapor signals, protein spectra normalization and secondary derivative analysis.
All supplies were from Invitrogen. Samples were diluted with NuPAGE LDS nonreducing 4x sampled buffer to a protein concentration of 0.05mg/mL and heated at 70°C for 10 min. Samples (1µg per lane) and SeeBlue Plus 2 molecular weight marker were applied onto 4-12% Bis-Tris pre-cast 1.0 mm x 10 well gel with 1x MOPS running buffer in a Mini-Cell. Electrophoresis was performed at 125 volts for 1.75 h. Gel was stained with Simply Blue Safestain, scanned via Amersham Biosciences Image scanner and quantified using ImageQuant ® software.
Vial size and FDS dilution
Initial attempts to shorten the reconstitution by adding salt or surfactant to WFI were unsuccessful (data not shown), hence other strategies were contemplated. First, using a larger vial (hence a larger surface area) should reduce the reconstitution time. Second, diluting the FDS prior to freeze-drying, as Shire et al (1) showed that decreasing protein loading concentration resulted in a less dense cake that reconstituted more readily. When six groups of samples with various vial sizes and FDS dilutions were freeze-dried in one cycle (Table II), the cake from the undiluted FDS reconstituted faster when placed in a 50mLvial (4.6 min) than in a 20 mL vial (6.4 min). Furthermore, the cake from the diluted FDS reconstituted significantly faster than that from the undiluted FDS, regardless of the vial size employed.
Table II. Effect of Vial Size and Dilution on Reconstitution
|Vial Size||Dilution||FDS||WFI||Reconstitution time|
|mL||mL||min ± SD||%|
|20 mL||None||5.5||0||6.4 ± 1.8||(100)|
|n=5||1.5-fold||5.5||2.75||4.0a ± 0.6||63|
|2-fold||5.5||5.5||3.1a ± 0.5||48|
|50 mL||None||5.5||0||4.6 ± 0.3||(100)|
|n=3||1.5-fold||5.5||2.75||3.8a ± 0.6||83|
|2-fold||5.5||5.5||3.0a ± 0.4||65|
|ap<0.05 compared to the undiluted controls, Student's t-test|
As shown in Figure 2, cycles with a 3 h annealing at either -3°C or -8°C had a significantly higher drying rate compared to annealing at -13°C or -18°C. However, -8°C was chosen over -3°C as the target annealing temperature to provide for a margin ensuring the temperature was well below the ice melting point.
When a formulation freezes, a phase separation occurs. A pure crystalline phase will separate from a saturated amorphous phase. The crystalline phase includes ice or any other crystallizing excipients (for example, sodium phosphate). During primary drying, the pure ice phase is removed, leaving behind other crystalline phases and any saturated amorphous phases. The aim of primary drying is to remove this unbound water while maintaining the cake structure and protein stability.
To prevent the cake from collapsing during primary drying, the product temperature must be kept under the collapse temperature (Tc). Because Tc is usually 2-3°C above glass transition temperature (Tg’), using the Tg’ to gauge the allowable product temperature represents a more conservative approach. As the Tg’ of the 25 mg/mL FDS was -20°C (data not shown), a maximum allowable product temperature of -25°C was selected to provide a 5°C safety margin during primary drying.
Two primary drying temperatures were examined: 10°C and 22°C. Primary drying at 10°C represented the more conservative viable cycle, while primary drying at 22°C was chosen to decrease the drying time. The cycle performed with a shelf temperature of 10°C and 22°C had a product temperature of -26°C and -21°C, respectively. Based on the proximity of the product temperature (-21°C) to the observed Tg’ (-20°C), the 22°C primary drying shelf temperature was not appropriate.
C. Secondary drying
Once the unbound, pure ice phase is removed, primary drying concludes. The next step is to remove water trapped in the amorphous phase during secondary drying. Because the Tg’ increases when primary drying is complete, the shelf temperature can be increased while keeping the product temperature below the Tg’. Increasing the shelf temperature increases the heat available to remove the bound water and hence increases the drying rate.
Two secondary drying temperatures were examined; 25°C and 40°C. The 25°C represented the more conservative viable step, while the 40°C was expected to decrease the drying time. Product stability, moisture content and cake appearance were used to evaluate the feasibility of the two different process conditions. Secondary drying at both temperatures had no impact on the target moisture content (of less than 1%), reconstitution time, pH or turbidity. In addition, the stability of the products generated from the two temperatures was not significantly different. DP form both temperatures lost 0.2% native protein compared to the pre-lyophilized FDS in the SE-HPLC assay, the most sensitive stability indicator for the DP. Therefore, a secondary drying temperature of 40°C was chosen to shorten the overall length of the cycle.
D. Comparison of the original and the revised cycle
The optimized revised cycle had a cycle length comparable to the original cycle. The original cycle with 5.5 mL fill at the primary and secondary drying temperatures of -5 °C and 25°C, respectively, was completed in about 50 h. Despite the larger volume and insertion of the optimized annealing step, the revised cycle with 8.8 mL fill at the primary and secondary drying temperatures of 10°C and 40°C, respectively, could be completed in about 52 h.
Reconstitution time: Effects of dilution, annealing and reconstitution method
Reconstitution times or dissolution rates depended on reconstitution method, and for the same reconstitution method, the rates were affected by the processing of the DP. All the 8.8 mL-filled tall cakes dissolved more rapidly than the original 5.5 mL-filled short cakes. Employing various methods including the swirling method, a mechanical orbital shaker rotating at 200rpm, or with the vial remained stationary after WFI addition (i.e., no agitation), the tall cake reconstituted significantly faster by 47%, 41% and 72%, respectively, compared to the short cake (Figure 4, A-C). If reconstitution was performed more forcefully with shaking by hand, the tall cake reconstituted in less than half a minute, or ¼ of the time required for the short cake (Figure 4D). Most interestingly, the tall cake not only reduced the reconstitution time but also reduced the variability in the reconstitution times compared to that of the short cake., The reduced variability is observed in a lower standard deviation (6 s versus 50 s) and smaller difference observed between the maximum and minimum reconstitution times (25 s versus 151 s) with n=20 (Figure 4D).
The above comparison revealed the combined effects of dilution and annealing. To explore the effect of the annealing per se on the reconstitution time, two groups of 8.8 mL-filled vials were lyophilized with and without the annealing step. All annealed tall cakes dissolved more rapidly than their non-annealed counterparts. Insertion of the optimized annealing step significantly reduced the reconstitution time by 39% and 45% with the swirling and shaking methods, respectively (Figure 5).
Figure 4. Comparison of reconstituting the 5.5 mL-filled versus 8.8 mL-filled cakes.
Figure 5. Effect of annealing on reconstitution time. Vials were all 8.8 mL-filled and produced with or without annealing (w/o A).
Characterization of the DP from the revised cycle
Besides reconstitution time, the physical attributes including cake appearance, moisture and turbidity of reconstituted solution of the lyophilized DP were examined.
No cracks, no collapse and no meltback of the cake were observed. The moisture content of the cake was less than 0.1%. Turbidity measured at 405 nm passed the specification indicating no significant particulates were produced during the lyophilization and reconstitution processes.
Concerns on shaking reconstitution method
Upon the more forceful shaking reconstitution, a significant amount of foam was visible at the top of the product, which subsided after about 10 min. However, this reconstitution characteristic did not impact the quality of the drug product. Several analytical methods demonstrated that the quality of the DP was maintained when reconstituted by the swirling or shaking methods. Reconstitution by the shaking method had no detrimental effect on either the recovery of the protein or the purity as determined by SE-HPLC. Figure 6A showed an overlay of SE-HPLC chromatograms of pre-lyophilized FDS and DP reconstituted by the swirling and shaking methods. Reconstitution by the shaking method had no effect on the purity of the protein. FTIR analysis showed no significant alterations in the secondary structure of the protein caused by a reconstitution method (Figure 6B).
In addition, the effects of swirling and shaking reconstitution methods on the integrity of the DP were examined by SDS-PAGE, forced degradation study and functional assay (potency or % bioassay). Reconstitution by the shaking method had no adverse effect on the integrity of the protein as determined by these assays. The formation of molecular weight species is the major degradation pathways of the DP. The DP does not oxidize or deamidate. A scan of the SDS-PAGE analysis confirms that reconstitution via shaking had no detrimental effect on the stability of the DP (Figure 6C). A forced degradation study examining the stability of the reconstituted DP after storage at 25°C (data not shown) and 37°C (Figure 6D) showed that the degradation profiles were identical for both methods. The acceptance criteria for the bioassay test are ± 50% of the reference standard (100%). The measured bioactivity was 115% and 91% for the swirling and shaking reconstituted DP, respectively.
Figure 6. Quality of reconstituted DP. A. SE-HPLC overlay. Sample traced in: Red, Pre-lyophilized FDS; Green, DP reconstituted by swirling; Blue, DP reconstituted by shaking. B. FTIR secondary derivative overlay. Sample traced in: Purple, Pre- lyophilized FDS; Green, DP reconstituted by swirling; Red, DP reconstituted by shaking. C. Non-Reducing SDS-PAGE analysis. Lane1: Pre-lyophilized FDS; Lane 2: DP reconstituted by shaking; Lane 3: DP reconstituted by swirling. D. SE-HPLC analysis for % native after storing the reconstituted DP (n=2 for each method) at 37°C for up to 15 days.
To reduce the reconstitution times encountered with highly concentrated protein pharmaceuticals, Shire et al (1) illustrated dilution as a viable approach provided the accompanied longer cycle time can also be reduced (1). Two key approaches were employed here to reduce the cycle time. One approach is to raise the drying temperatures, and the other is to add an annealing step (2). While raising the drying temperatures is relatively straightforward, adding an annealing step is not, especially because literatures on the effects of annealing are diverse at best.
Annealing involves holding the frozen product above the Tg’ of the formulation prior to the initiation of primary drying. Such a step potentially accomplishes two things: 1. it allows for crystal growth of excipients that did not fully crystallize during the initial freezing step; 2. it potentially reduces the freezing induced heterogeneity in drying rate through Oswald ripening. By raising the temperature of a vial above the Tg’, ice crystals are allowed to reorganize to a lower energy state, thereby increasing the size of large crystals at the expense of eliminating smaller ones. In this case, formation of larger crystals minimizes the surface area which reduces the total energy of the system.
Depending on the formulation and process variables, the above two consequences of annealing have opposite effects on the primary drying rate. Searles et al (3) noted that annealing could increase the primary drying rate for a model compound hydroxyethyl starch by allowing ice crystals grow bigger (through an Ostward ripening process), hence leaving larger pores or channels for water vapor flow in the already-dried layer. Similarly, Webb et al (4) also noted that annealing reduced the overall length of the lyophilization cycle for the recombinant human interferon-ɤ. However, Lu and Pikal (5) showed that annealing could slow down the primary drying when annealing caused crystal growth of certain excipients that blocked the water vapor flow and hence increased the dry layer resistance.
To accurately determine the effects of annealing on the rate of drying in the present study, all samples needed to undergo primary drying in the same cycle. This was accomplished by storing the samples in a -40°C freezer after the individual annealing step had been performed. Upon completion of the various annealing treatments, the vials including the controls were transferred to the -40°C shelf in the lyophilizer prior to initiate primary drying.
Besides influencing primary drying rate on a case by case basis, annealing has been reported to either decrease or increase dissolution rate on a case by case basis. Searles et al (3) reported the annealed samples dissolved slightly faster than their unannealed counterparts. On the contrary, Webb et al (4) noted annealing caused an 18-fold reduction in the dissolution rate of lyophilized interferon-r through a reduction in the surface area of the cake available for wetting.
In the current report, the insertion of the optimized annealing step significantly increased not only the primary drying rate, but also the dissolution rate of the lyophilized DP (Figure 5).
In fewer than 3 months and before the Phase III clinical trial DP was being produced, the recommended changes as reported here was sent to the contract manufacture site for implementation.
Implementing dilution, annealing and shaking method reduced not only the reconstitution time to fewer than one min, but reconstitution time heterogeneity among vials. The lyophilization cycle was also modified to accommodate the large fill volumes by adding the annealing step and by increasing the primary and secondary drying temperatures. These changes did not significantly compromise the DP quality nor the cycle duration. Hence these changes have been incorporated into the manufacture and reconstitution of the Phase III clinical trial material, and eventually the commercial DP.
Dr. Leu-Fen H. Lin
Senior Manager, Formulation Development
SGS Life Science Services
Dr. Richard Bunnell
SGS Life Science Services
The authors thank Drs. Katherine Bowers and Karen Bossert for suggesting the shaking reconstitution, and Abby Thummals for excellent technical support.
- S.J. Shire, Z. Shahrokh and J. Liu, J Pharm Sci. 93, 1390-1402 (2004).
- M.J. Pikal, BioPharm. 3, 18-27 (1990).
- J.A. Searles, J.F. Carpenter and T.W. Randolph, J Pharm Sci. 90, 872-887 (2001).
- S.D Webb, J.L. Cleland, J.F. Carpenter and T.W. Randolph, J. Pharma Sci. 92, 715-729 (2003).
- X. Lu and M.J. Pikal, Pharm Dev & Tech. 9, 85-95 (2004).