The Impact of Drying Method and Formulation on the Physical Properties and Stability of Methionyl Human Growth Hormone in the Amorphous Solid State

Transcription

1 BIOTECHNOLOGY The Impact of Drying Method and Formulation on the Physical Properties and Stability of Methionyl Human Growth Hormone in the Amorphous Solid State AHMAD M. ABDUL-FATTAH, 1 DAVID LECHUGA-BALLESTEROS, 2 DEVENDRA S. KALONIA, 1 MICHAEL J. PIKAL 1 1 Department of Pharmaceutical Sciences, University of Connecticut, Storrs, Connecticut Nektar Therapeutic Systems, San Carlos, California Received 10 October 2006; revised 5 January 2007; accepted 22 January 2007 Published online in Wiley InterScience ( DOI /jps ABSTRACT: The objective of this work was to investigate the impact of drying method and formulation on the physical stability (aggregation) and selected important physical properties of dried methionyl human growth hormone (Met-hGH) formulations. Solutions of Met-hGH with different stabilizers were dried by different methods (freeze drying, spray drying, and film drying), with and without surfactant. Properties of the dried powders included powder morphology, specific surface area (SSA), protein surface coverage, thermal analysis, and protein secondary structure. Storage stability of MethGH in different formulations was also studied at 508C and at 608C for 3 months. The dried powders displayed different morphologies, depending mainly on the method of drying and on the presence or absence of surfactant. Film dried powders had the lowest SSA (0.03 m 2 /g) and the lowest total protein surface accumulation (0.003%). Surfactant caused a reduction in the SSA of both spray dried and freeze dried powders. Spray dried powders showed greater protein surface coverage and SSA relative to the same formulations dried by other means. Greater in-process perturbations of protein secondary structure were observed with polymer excipients. Formulation impacted physical stability. In general, low molecular weight stabilizers provided better stability. For example, the aggregation rate at 508C of Met-hGH in a freeze dried trehalose-based formulation was approximately four times smaller than the corresponding Ficoll-70- based formulation. Drying method also influenced physical stability. In general, the film dried preparations studied showed superior stability to preparations dried by other methods, especially those formulations employing low molecular weight stabilizers. ß 2007 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 97: , 2008 Keywords: freeze drying; spray drying; film drying; human growth hormone (hgh); composition heterogeneity; electron spectroscopy for chemical analysis (ESCA); glass transition temperature (T g ); particle morphology; protein aggregation; protein secondary structure; specific surface area Ahmad M. Abdul-Fattah s present address is Drug Product Device Development, Amgen, Thousand Oaks, CA Correspondence to: Michael J. Pikal (Telephone: ; Fax: ; michael.pikal@uconn.edu) Journal of Pharmaceutical Sciences, Vol. 97, (2008) ß 2007 Wiley-Liss, Inc. and the American Pharmacists Association INTRODUCTION Freeze drying is widely used for the preparation of protein pharmaceuticals for parenteral administration, 1 3 while spray drying is widely used for pulmonary delivery of protein pharmaceuticals. 4,5 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY

2 164 ABDUL-FATTAH ET AL. Other drying methods, such as spray coating, 6 supercritical fluid technology, 7 and spray freeze drying 8 have also been used to dry protein solutions. Drying by different methods subjects the protein to different thermal histories and different stresses, resulting in potential differences in physical properties, such as moisture content, particle size, particle morphology, powder density, specific surface area (SSA), surface composition, and thermal properties. 6,9 12 It is becoming accepted 9,13 19 that there may be significant separation of stabilizer from protein during drying without actual phase separation. That is, the surface region becomes rich in protein, and the interior therefore becomes richer in stabilizer, resulting in an uneven distribution of chemical components throughout the dried particle. 9,13 19 Such a phenomenon is referred to as composition heterogeneity or chemical heterogeneity. Phase separation, on the other hand, refers to the formation of two or more distinct phases. Examples of phase separation include the crystallization of one or more components (such as a buffer component and/or a bulking agent) from the amorphous phase, 20,21 or the formation of two different amorphous phases (as has been documented with PEG/dextran mixtures). 22 Composition heterogeneity has been documented using special surface analysis techniques, such as electron spectroscopy for chemical analysis (ESCA). 9,13 19 Phase separation in amorphous systems has been documented by techniques such as electron microscopy and differential scanning calorimetry. 25,26 Composition heterogeneity perhaps occurs as a result of differences in diffusion coefficient between formulation components in solution and/or because of differences in surfactant properties of the components. 9,13 19 As might be expected, the inclusion of a surfactant in the formulation significantly moderates the accumulation of protein at the surface of spray dried samples. 9,13,14,17 19 Development of composition heterogeneity from a solution initially uniform in composition clearly requires molecular mobility. That is, the system must possess sufficient molecular mobility to support the mutual diffusion needed to create separation of components. Composition heterogeneity could arise during freezing for a freeze drying process and during much of the drying process for a spray dried material. However, once the material is well below the glass transition temperature, it seems unlikely that sufficient translational mobility would exist JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 to allow further separation of components on the time scale of the process. Moreover, although there is spatial variation in composition in freeze dried materials, it does appear that composition variation in spray dried materials is typically an order of magnitude larger. 9,18 Since one normally finds that the stability of a given protein improves as the weight ratio of stabilizer to protein increases, it follows that the separation of components might have adverse stability consequences. That is, the protein near the surface would be much less stable than what the overall composition would suggest, and the protein in the interior may not gain sufficient stability to compensate. However, the implications of such component separation on storage stability are poorly understood. It is becoming well known that stability of proteins in amorphous pharmaceuticals shows a log-linear dependence on composition Assuming that the measured or observed rate constant for degradation (k obs ) is a summation of contribution of stability from both surface and bulk protein, we may write: k obs ¼ k S F PS þ k B F PB (1) where k S is the rate constant for decomposition of surface protein, k B is the rate constant for decomposition of bulk protein, F PS is the fraction of the surface protein and F PB is the fraction of bulk protein. Note that surface protein as used in Eq. (1) refers to the fraction of total protein which is within the first 50 Å outer shell thickness, as measured by ESCA. As described by Eq. (1), an increase in surface protein increases thefirstterm on the right hand side of Eq. (1) and decreases the magnitude of the second term. Numerical calculations show that when the rate constant shows a loglinear dependence on composition, the net effect of composition heterogeneity is an increase in the magnitude of k obs. That is, the net effect decreases stability relative to what one would expect if no separation of components occurred. This effect is expected to be more significant for formulations rich in stabilizer. 32 A few studies have shown that formulations with a high SSA had inferior storage stability relative to the same formulations with lower SSA, as prepared by other drying methods. 9,33 For example, Sane et al. 33 observed that the storage stability of a freeze dried antibody formulation was superior to the stability of the same formulation prepared by spray drying. Protein secondary structure was the same in both formulations, but the SSA of the spray dried formulation was DOI /jps

3 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE 165 higher. Webb et al. 9 compared selected solid state properties and storage stability of spray freeze dried and freeze dried trehalose formulations of recombinant human interferon-g (rhifn-g) in the presence and absence of Tween-20. The spray freeze dried formulation without Tween-20 had the highest SSA and also had the highest aggregation rate constant. Second derivative FTIR spectra showed that protein secondary structure was similar in all preparations. The lyophilized formulation with Tween-20 had the lowest SSA and lowest aggregation rate constant. We suggest that composition heterogeneity phenomenon played a major role in determining the relative stability of material produced by the two processes. This study addresses the impact of drying method and formulation (i.e., type of stabilizer and the presence of surfactant) on the solid state properties of the dried material. Specifically, we studied the impact of drying method and formulation on composition heterogeneity, protein secondary structure, and storage stability (aggregation) in the solid state. In our studies, methionyl human growth hormone (Met-hGH) was used as the model protein. Met-hGH differs from hgh in that it has 192 amino acid residues. 34,35 For most assays, as well as physical and chemical properties, the presence of the NH 2 terminal methionine in Met-hGH is unimportant. 36 Stabilizers used in this study were either the low molecular weight saccharides, trehalose (disaccharide) and stachyose (tetra-saccharide), or polysaccharides (dextran 40 and Ficoll 70). MATERIALS AND METHODS Materials Recombinant Met-hGH with mannitol and glycine (1:5:1 by weight) lyophilized from 5 mm Na 2 HPO 4 (ph 7.6) was obtained from BresaGen (Adelaide, Australia). Chemicals and excipients were all used as supplied. Trehalose (mol. wt. 342), stachyose (mol. wt ), Tween-20 (monomer mol. wt. 1228), Ficoll 70 (average mol. wt ), potassium phosphate dibasic, and potassium phosphate monobasic were purchased from Sigma-Aldrich (St. Louis, MO). Dextran 40 (average mol. wt ) was purchased from ICN Biomedicals (Aurora, OH). Twenty millimeters of 20 ml type I borosilicate clear tubing glass vials, as well as V10 F mm single-vent Flurotec 1 (Daikyo Seiko) lyophilization stoppers were purchased from West Pharmaceuticals (Lionville, PA). Preparation of Met-hGH Solutions All solutions were prepared by dialysis of an aqueous solution of Met-hGH (2 mg/ml) against phosphate buffer (sodium phosphate 0.15 mg/ml, 1.1 mm) ph 7.8 with stabilizer (6 mg/ml). The solutions were prepared with and without Tween- 20 (1 mg/ml). Solutions were dialyzed for 15 h at 48C and dialysis buffer was changed once during the course of dialysis. Total solids content in the final dialyzed solutions was less than 1% w/v. Surface Tension Measurements Static surface tension of all solutions was measured using a Krüss K121 Contact Angle and Adsorption Measuring System. Measurements were performed on ml of the solutions at room temperature in duplicate. Drying of Met-hGH Solutions Spray Drying Dialyzed Met-hGH solutions were spray dried with a Buchi 190 Mini Spray Drier (Buchi, Flawil, Switzerland). The solution feed rate was 5 ml/ min. Inlet temperature ranged from 91 to 948C, outlet temperature ranged from 53 to 588C and atomization pressure was 40 psi. Moisture content in spray dried samples ranged between 4% and 5% w/w after spray drying, as measured by Karl Fischer moisture titration. The target moisture content in all dried formulations was less than 1% w/w. Hence, all spray dried lots were subjected to a vacuum drying step after spray drying. Vacuum drying was performed in a FTS Systems Dura Stop Dryer in conjunction with a Dura Dry MP freeze dryer (Stone Ridge, NY) at a shelf temperature of 388C and a chamber pressure of 100 mtorr for 9 h in 20 ml glass vials. Vials were sealed under vacuum at the end of the drying cycle. Freeze Drying Dialyzed Met-hGH solutions were freeze dried using an FTS Systems Dura Stop Dryer in conjunction with a Dura Dry MP freeze dryer. Solutions were lyophilized in 20 ml serum vials with a fill volume of 2 5 ml. Solutions were first frozen by reducing the shelf temperature to 408C DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008

4 166 ABDUL-FATTAH ET AL. and holding at 408C for 3 h followed by primary drying well below the collapse temperature (lowest glass transition temperature, T g 0, for freeze concentrate recorded by MDSC was 348C). Shelf temperature was held at 348C for 37 h during primary drying while maintaining a chamber pressure of 50 mtorr. Average product temperature during primary drying was approximately 378C. The end of primary drying was marked by a rise in product temperature above the shelf temperature with a concurrent decrease in dew point to a constant value. Secondary drying was done at 338C for 5 h at 100 mtorr. Cakes produced after freeze drying were elegant with excellent retention of cake structure. Freeze Drying with Annealing Some Met-hGH samples were annealed in the frozen state prior to primary drying to produce samples of lower surface area and to determine if annealing damages stability. Annealing was performed at a shelf temperature of 258C for 35 h. Shelf temperature was lowered again to 408C and samples held to equilibrate at this temperature for 1.5 h prior to primary drying. Conditions for primary and secondary are the same as described under Freeze drying. Cakes produced after freeze drying were elegant with excellent retention of cake structure. Film Drying Film drying was mainly performed to produce samples with very low surface area without freezing. Film drying, in the context of this work, is a form of controlled evaporative drying, as addressed by Franks et al., 37 that produces a glassy dry state without formation of ice during the process. The method has also been referred to as liquid film dehydration by Schroederand and Schwarz. 38 In either case, the system is dried at very low pressure, and heat input into the material is adjusted to achieve a high rate of evaporation. The equipment necessary to perform film drying consists of a drying chamber, a condenser and a vacuum pump. 39,40 We used a conventional freeze dryer to film dry the solutions. Dialyzed Met-hGH solutions were film dried using a FTS Systems Dura Stop Dryer in conjunction with a Dura Dry MP freeze dryer in 20 ml serum vials. Shelf temperature was held at 378C for 2 h while maintaining chamber pressure at 3000 mtorr during the initial drying phase. Under these conditions, the solutions JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 neither froze, boiled nor foamed. Instead, all solutions dried into a film without freezing within approximately 1 h. Product temperature did not go below 28C during the primary drying phase. Secondary drying was performed by heating the shelf temperature to 338C at a rate of 18C/min and the films were held at that temperature for 5 h at 100 mtorr. After secondary drying, the product was crushed in the vials with a spatula into small flakes. Initial Karl-Fischer titration indicated a moisture level above 2% w/w in the film dried product. Further vacuum drying was performed at 388C and a chamber pressure of 100 mtorr for 9 h to reduce moisture content of the samples to less than 1% w/w. Karl Fischer Moisture Determination Moisture content in the dried powders was measured by Karl Fischer titration by direct injection using a Mitsubishi model CA-06 moisture meter (Mitsubishi Kasei Corporation, Tokyo, Japan). Powders were dissolved/dispersed in 2.5 ml low moisture formamide and 1 ml of the solution was injected. Titration checks were made by injecting water standards (Mitsubishi MC02020) at several time intervals between injections. Blank corrections were applied. Standard deviation from replicate measurements was not greater than 0.1. Scanning Electron Microscopy (SEM) Studies SEM analysis was performed to characterize morphology of spray dried, freeze dried, and film dried products. Samples were mounted on silicon wafers mounted on top of double-sided carbon tape on an aluminum SEM stub. The sample was then sputter coated in a Denton sputter coater for s at 75 mtorr and 42 ma with gold/ palladium. Images were taken with a Philips XL30 ESEM operated in high vacuum mode using an Everhart Thornley detector to capture secondary electrons for the image composition. Accelerating voltage was set at 20 kv using a LaB 6 source. Working distance was 5 6 mm. Specific Surface Area (SSA) Determination SSAs were measured using a Micromeritics Gemini III 2375 BET nitrogen adsorption surface area analyzer (Norcross, GA). Sample size was at least 100 mg. Powder samples were degassed for DOI /jps

5 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE h in the Flow Prep oven at 358C using N 2 purge prior to measurement. Multiple point (Brunauer, Emmett, and Teller: BET) analysis was performed using five points in the range of for P/P 0 (relative pressure). To evaluate the reproducibility of BET SSA measurements, replicate measurements (n ¼ 5) were carried out on a spray dried sample of MethGH. The mean SSA from multiple point analysis in that sample was m 2 /g (i.e., relative standard deviation (RSD) was less than 0.5%). Electron Spectroscopy for Chemical Analysis (ESCA) Studies ESCA was used to determine protein surface concentration in all dried formulations. Single samples of each powder, each weighing approximately 10 mg, were processed and analyzed by ESCA. A glass plate cleaned with 10% hydrofluoric acid solution was used to press each powder sample on a glass substrate that had been sandblasted to improve the adhesion of the powder to the glass. A monochromatic Al X-ray ( ev, 300 W) with a known take-off angle of 458 was directed to a 2 3 mm spot on the sample, and an electron flood gun operating at 2.5 ev was used for neutralization. Survey (pass energy ¼ ev) and high-resolution (pass energy ¼ ev) spectra were collected with a Physical Electronics PHI 5000 Series Spectrometer, and the peak areas from the high-resolution spectra were converted to elemental atomic concentrations on the surface of each sample. To examine the reproducibility and machineto-machine variability for ESCA measurements, survey spectra were also collected with a VG ESCALAB MK II series Spectrometer (pass energy ¼ 60 ev). An Al ka radiation (emission current of 11 kv anode voltage and 34 ma emission current) from Al Ka X-ray source was directed onto a circular region of approximate diameter of 3 mm. Pressure was maintained in analysis chamber at less than 10 8 Torr. The mean difference between measurements was 4.8% (Tab. 1). Data Analysis Each component in the powder is characterized by the specific ratio between its elements. From an analysis of the relative amount of the different elements in the pure components and in the powder, one can estimate the percentage of powder surface covered with each chemical component. Met-hGH mainly contains carbon (C), oxygen (O), and nitrogen (N). Met-hGH contains approximately 16.8 wt% N. ESCA measurements gave wt% N for pure Met-hGH. Met-hGH also contains 0.39 wt% S, an amount below the detection limits of ESCA (at least 0.5 wt% required per manufacturer s recommendations). Both the stabilizer and surfactant contain only C and O. Thus, the N peak is indicative of the presence of Met-hGH in the outer 50 Å surface 9,16 of a sample. Protein surface coverage, 13 15,17 19 defined as the wt% of Met-hGH that exists in the surface region (50 Å from surface), was determined via calculations presented in Appendix 1. 15,16,32,41 Formulations containing surfactant are loaded with 21.5% w/w Met-hGH. So it follows that the surface layer should contain approximately 3.66 wt% N or 21.5% w/w Met-hGH if Met-hGH were homogeneously distributed throughout. Similarly, surfactant-free formulations are loaded with 24% w/w Met-hGH, hence the surface layer should contain approximately 4.1 wt% N or 24% w/w of Met-hGH if Met-hGH were homogeneously distributed. The ratio of protein surface coverage (% w/w; measured by ESCA) to the surface coverage (% w/w) of a hypothetical homogeneous MethGH powder is referred to as surface coverage ratio in the current study. Surface coverage ratio is an indication of surface excess or surface deficiency of protein in the surface region. 13,14 For the Table 1. Variability in ESCA Measurements with Protein Surface Concentration of Different Preparations of Met-hGH Using Two Different Instruments Stabilizer Tween-20 Drying Process Weight % of Met-hGH at Surface (Physical Electronics PHI 5000 Series) Weight % of Met-hGH at Surface (VG ESCALAB MK II Series) Trehalose Y Film dried 5 7 Stachyose Y Freeze dried 8 15 Dextran 40 N Freeze dried Ficoll 70 Y Freeze dried 5 8 N Freeze dried DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008

6 168 ABDUL-FATTAH ET AL. current formulations, surface excess means greater than 24% w/w Met-hGH on the surface of surfactant-free dried formulations, and greater than 21.5% w/w Met-hGH on the surface of dried formulations with Tween-20. Surface deficiency means less than 24% w/w Met-hGH on the surface of surfactant-free dried formulations, and less than 21.5% w/w Met-hGH on the surface of dried formulations with Tween-20. Neither protein surface coverage nor surface coverage ratio account for differences in SSA between different dried preparations. Using protein surface coverage, the volume fraction of the powder sensed by ESCA, and SSA of the powder, the total protein surface accumulation (%P total-surface ) was determined, that is, how much of the total protein (% w/w) in the formulation has accumulated at the surface. %P total-surface represents the protein fraction in a potentially reactive state. Details for calculations to determine %P total-surface are discussed in Appendix 1. 32,41 Modulated Differential Scanning Calorimetry (MDSC) Studies MDSC studies were performed using a TA 2920 instrument (TA Instruments, New Castle, DE) to determine the glass transition temperature (T g )of the different dried formulations, primarily to aid in proper design of storage stability studies. A mass of 5 15 mg of the powdered sample was filled and pressed into compact pellets in aluminum DSC pans and hermetically sealed using a sample encapsulation press. The entire process was performed in a glove box, with RH maintained under 2%. Helium was used as a purge gas at 30 cm 3 /min; the Refrigerated Control System (RCS) used helium at a flow rate of 110 cm 3 /min. All MDSC measurements were done in triplicate. Ramp rate was 28C/min and modulation was C every 60 s. Fourier Transform Infra Red (FT-IR) Spectroscopy Studies FTIR spectroscopy studies were performed using a Nicolet Magna IR 760 Spectrometer (Thermo- Nicolet, Madison, WI). FTIR studies were performed on Met-hGH in solution (i.e., native conformation), approximately mg/ml, and in selected dried solid formulations according to procedures in the literature A total of 256 scans and a resolution of 4 cm 1 were used for each spectrum. The scanning range was 4000 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY cm 1. The spectrum for native, aqueous MethGH (used as a control) was obtained by placing a 15 mg/ml solution between 2 CaF 2 windows separated by a 6 mm Mylar spacer. The spectrum of Met-hGH in solution was subtracted from the spectra of liquid water and water vapor as described elsewhere. 43 The solid samples were prepared by pressing a ground homogeneous mixture of 100 mg KBr with 2 3 mg of the dried protein formulation into a pellet (disk) using a stainless steel die. Effects of water vapor and carbon dioxide in the chamber were eliminated by subtracting the appropriate background spectrum. The second derivative IR spectra of the samples were obtained using Omnic software version 6.0a 1 (Nicolet, Madison, WI). The amide I region ( cm 1 ) of the spectra (smoothed by seven point smoothing) were baseline corrected and area normalized, then overlaid to compare the spectra. Conformational changes of Met-hGH in dried samples were characterized by comparing secondary structure of Met-hGH in the amide I region with the solution spectra. The correlation coefficient (r) was determined to compare the similarity between spectra of the dried formulations and that of aqueous solution, 48 defined by the following formula: P ðxi y i Þ r ¼ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P P (2) x 2 i y 2 i where x i and y i are the intensity of the spectra for solid samples and the spectra of protein in solution (native structure) at wave number (i), respectively. RSD for determination of r from FTIR studies (replicate samples) did not exceed 2.4%. Stability Studies and Assays Physical degradation of Met-hGH involves aggregation by covalent or noncovalent bonds, usually noncovalent. Aggregates may be water soluble or insoluble, insoluble aggregates leading to turbid solutions. 34 HPLC Assay Methodology Total covalent and noncovalent hgh soluble aggregates were measured by size exclusion chromatography (SEC). The liquid chromatography (LC) system consisted of a Waters 515 HPLC pump equipped with a Spectra system UV1000 multiple wavelength detector, a Waters 717 plus automatic sampling system and a Waters In-line DOI /jps

7 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE 169 degasser. Integration was done with Peak Simple software version The LC column was a TosoHaas TSK Gel 2000SWxL column (5 m, 125 Å, mm ID). The mobile phase consisted of 3:97 (v/v) isopropanol-63 mm potassium phosphate dibasic aqueous buffer (ph 7 adjusted with o-phosphoric acid). The method was run isocratic at 1 ml/min. The wavelength for detection was 214 nm, the column temperature was ambient and the injection volume was 50 ml. Samples were run in triplicate. Stability Studies and Characterization of Degradation Rates Selected Met-hGH formulations were stored in stability ovens at 50 18C and at 60 18C. The samples were analyzed by the SEC HPLC method initially and after 1 and 3 months. At least two samples were assayed from each lot at each time point and replicate injections for each sample were made. Purity level determination by the SEC method was tested for reproducibility (inter-vial variability and long-time drift) and precision. % RSD for the precision test (multiple injections from the same vial) was not greater than 0.1%. %RSD for inter-vial variability was not greater than 1.1%. Additionally, inter-month variability (i.e., longtime drift) was assessed by storing two samples with different initial purity (treated as control samples) at 208C for 1 and 3 months after which the samples were reassayed (samples were labeled A and B ). Initial purities were and for samples A and B, respectively. At the 1 month assay time the corresponding values were and At 3 months the values were and That is, % RSD for long-time drift was 1.6% for sample A and 0.3% for the higher purity sample, sample B. RESULTS AND DISCUSSION Effect of Protein and Surfactant on Surface Tension in Solution As expected and as is consistent with literature, 49 Met-hGH significantly reduced the surface tension of surfactant-free solutions (Tab. 2) because of its surface activity. A lower surface tension was observed in Met-hGH-free solutions with 0.1% w/v Tween-20. It is known that surfactants preferentially adsorb at air/water interfaces thereby minimizing protein aggregation in solution by reducing protein adsorption and unfolding at the interface. 50 No significant reduction was observed in the surface tension of solutions containing Tween-20 upon the addition of Met-hGH. Hence it is likely that Tween-20 competes effectively with Met-hGH for the air/water interface and largely excludes the protein from the surface. Moisture Content and Glass Transition All dried Met-hGH preparations were devoid of Bragg peaks as revealed by XRPD studies (results not shown), suggesting essentially pure amorphous character. Final adjusted moisture content for all dried formulations was less than 1% w/w. Glass transition temperature (midpoint; T g ) for all preparations were well above 1008C (results not shown). Drying method did not have a significant impact on glass transition behavior. Table 2. Surface Tension Measurements Stabilizer in Solution Surfactant (1 mg/ml) Met-hGH (2 mg/ml) Surface Tension (Dynes/cm 2 ) None (DI distilled water) Trehalose þ þ þ þ Stachyose þ þ þ Dextran 40 þ þ þ Ficoll 70 þ þ þ þ DI, deionized water. DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008

8 170 ABDUL-FATTAH ET AL. Effect of Drying Method and Formulation on Particle Morphology Powders displayed different morphologies, as viewed by SEM, depending mainly on the method of drying and on the presence or absence of surfactant. In most cases, the type of excipient used had no significant influence on morphology as was also the case in studies by Millqvist-Fureby et al. 18 Freeze dried powders showed thin, leaf shaped flakes (Fig. 1A and B) for both surfactant-free and surfactant containing formulations. Annealing during freezing produced little change in morphology, but perhaps produced slightly larger flakes (Fig. 1B). Film dried samples were in the form of large thick plates (Fig. 1C) containing small pores (pores not visible at the magnification shown), suggesting possible bubble formation during the early part of drying. The effect of Tween-20 on particle morphology was most noticeable with spray dried powders. It is known that morphology of spray dried particles is affected by protein content; the more protein, the more wrinkles and dents on the surface In absence of Tween-20, particles were raisin-like showing multiple wrinkles and dents on the surface (Fig. 1D). In presence of Tween-20, particles were essentially spherical and smooth, but with a few dimples (Fig. 1E). In some fields of view, some fusion of particles is evident (figures not shown). Effect of Drying Method and Formulation on SSA and Surface Composition SSA SSA was impacted mainly by drying method and presence of surfactant (Tab. 3). In the absence and presence of surfactant, SSA was highest in spray dried formulations. In the presence of surfactant, film dried preparations had extremely low SSA. SSA increased in the following order: Film dried powders freeze dried powders < spray dried powders. The addition of Tween-20 caused a significant reduction (factor of 5) in the SSA of spray dried powders. High surface corrugation (i.e., roughness; Fig. 1D) is the likely cause of high SSA obtained with the surfactant-free spray dried powders. The addition of Tween-20 also caused a significant reduction in the SSA (about times) of freeze dried powders. Furthermore, annealing in the frozen state caused roughly a 1.6- to 2.3-fold reduction in the SSA of the freeze JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 dried powders. Annealing is a process by which samples are held at a temperature between the ice melt temperature and T g 0 for a period of time. Annealing allows larger ice crystals to grow at the expense of small ones in a process called Ostwald Ripening. 21 Below T g 0, the system is in a glassy state and ice crystal growth is limited. Above T g 0, growth is possible. The net result is an increase in ice crystal size, resulting in a product with a significantly lower SSA than the same formulation not annealed In systems with limited miscibility, phase separation and crystallization may also occur during annealing, but such complications would not be expected to be a factor in the systems studied in this research except for possible phase separation of surfactant. Protein Surface Coverage Figure 2A and B shows typical ESCA scans for dried powders with and without surfactant. Note that addition of surfactant greatly reduces the nitrogen peak, indicating much less protein at the surface with surfactant in the formulation. Protein surface concentration and surface coverage ratio are impacted mostly by drying method and the addition of surfactant (Tab. 3). Regardless of the presence of surfactant, protein surface coverage was highest with spray dried powders, and film dried preparations had the lowest protein surface coverage. Protein surface coverage increased in the following order: film dried freeze dried < spray dried. The addition of surfactant significantly lowered protein surface coverage in spray dried (factor of 3 4) and freeze dried powders (factor of 3 6). It is apparent that presence of surfactant prevented surface accumulation of protein, likely by displacing protein from the surface during the early part of drying. Thus, the composition at the air water interface during the early part of the drying processes determined the surface composition of the final powders In principle, composition at the aqueous-ice interface is impacted by the ability of molecules to freely diffuse during the process. Thus, the longer process time characteristic of annealing could also partially remove chemical heterogeneity. However, the data (Tab. 3) show no evidence for such an effect. We also note that with spray dried surfactantfree powders, protein surface coverage was higher with formulations containing higher molecular weight excipients than with trehalose (i.e., nearly DOI /jps

9 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE 171 Figure 1. Morphology of dried formulations as observed by SEM. Magnification for each image is given on the SEM panel. (A) Freeze dried Met-hGH/trehalose with Tween-20. (B) Freeze dried Met-hGH/trehalose with Tween-20 annealed prior to primary drying. (C) Film dried Met-hGH/Ficoll 70 with Tween-20. (D) Spray dried Met-hGH/trehalose without Tween-20. (E) Spray dried Met-hGH/trehalose with Tween-20. DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008

10 172 ABDUL-FATTAH ET AL. Table 3. Specific Surface Area (SSA), Protein Surface Coverage (% w/w as Determined by ESCA), Surface Coverage Ratio and Total Protein Surface Accumulation, %P total-surface, for All Dried Formulations (See Text for Details on Calculations) Formulation Treatment BET Surface Area (m 2 /g) Weight % of Met-hGH at Surface (Protein Surface Coverage) Surface Coverage Ratio %P total-surface a 1:3 (Met-hGH/trehalose) Freeze dried Spray dried :3 (Met-hGH/trehalose) Freeze dried with Tween-20 Annealed/ freeze dried Spray dried Film dried :3 (Met-hGH/stachyose) Freeze dried with Tween-20 Annealed/ freeze dried Spray dried Film dried :3 (Met-hGH/dextran 40) Freeze dried Spray dried :3 (Met-hGH/dextran 40) Freeze dried with Tween-20 Spray dried :3 (Met-hGH/Ficoll 70) Freeze dried :3 (Met-hGH/Ficoll 70) with Tween-20 Spray dried Freeze dried Spray dried Film dried a Standard error estimation calculated for %P total-surface based on propagation of errors was 1 for surfactant-free spray dried powders, 0.1 for surfactant-free freeze dried powders, and 0.01 preparations with surfactant. 60% compared to 40%, versus about 22% for a homogeneous system). If differential diffusion rates were the dominant factor in chemical heterogeneity, one would expect less separation of components in systems where the components have similar molecular weights and therefore similar diffusion coefficients However, it seems obvious that factors other than diffusion coefficient influence protein surface coverage, with surface activity being one additional factor. 9,13 19 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 Total Protein Surface Accumulation Total protein surface accumulation was impacted mainly by drying method and presence of surfactant. Regardless of the type of stabilizer and of the presence of surfactant, spray dried preparations showed the highest %P total-surface values, mostly as a result of their large SSA. That is, even when protein surface coverage in spray dried materials was comparable to freeze dried materials, %P total-surface was larger. On the other hand, film dried preparations had the lowest %P total-surface (in formulations with surfactant) as a direct result of their extremely low SSA. Annealing in the frozen state did not impact either protein surface coverage or %P total-surface,at least for the systems studied. Addition of surfactant caused greater than 10-fold reduction in %P total-surface in spray dried formulations and greater than 4-fold reduction in %P total-surface for freeze dried formulations, an effect caused by both the reduction in SSA upon addition of surfactant and the reduction of protein surface coverage. Effect of Drying Method and Formulation on Protein Secondary Structure in the Solid State Figure 3 compares FTIR spectra of Met-hGH in the amide I region ( cm 1 ) for spray dried formulations with the spectra of Met-hGH in DOI /jps

11 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE 173 Figure 2. Scans from ESCA studies plots of counts per second (CPS) versus binding energy (ev). (A) ESCA scan for Met-hGH/trehalose freeze dried formulation showing the main component atoms of the molecule (N, O, C). S atom peak is too small to be resolved. (B) Met-hGH/trehalose/Tween-20 freeze dried formulation showing a reduction in intensity of N peak indicative of less protein surface coverage. DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008

12 174 ABDUL-FATTAH ET AL. Figure 3. Second derivative FT-IR spectra at amide I region for some spray dried formulations: (1) Met-hGH solution, (2) Met-hGH with trehalose, (3, 4) Met-hGH with stachyose/tween-20 and Met-hGH with trehalose/tween20, (5) Met-hGH with dextran 40/Tween-20, (6) Met-hGH with Ficoll 70/Tween-20, (7) Met-hGH with Ficoll 70, and (8) Met-hGH with dextran 40. aqueous solution. FTIR spectra for freeze dried formulations were similar (results not shown). In solution, the FTIR spectrum of Met-hGH in the amide I region is dominated by a strong a-helix band at 1656 cm 1. 45,55 Upon drying, this band decreases in intensity and shifts to a higher wave number, accompanied by band broadening. Loss in intensity of a-helix band is accompanied by an increase in absorbance of b-sheet band at 1685 cm 1. Loss of native structure in the dried state showed a strong dependence on the stabilizer used. Lower molecular weight stabilizers (trehalose and stachyose) showed the greatest ability to maintain a sharp a-helix band in drying regardless of the drying method and regardless of the absence or presence of surfactant. Drying formulations with dextran 40 and Ficoll 70 resulted in a smaller and broader a-helix band and sharper and more intense bands at 1685 cm 1 (b-sheet band sheet) and 1635 cm 1. Similar results were obtained by Izutsu et al. 46 with BSA. Whereas incorporation of polymers such as dextran 40 or JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 Ficoll 70 can be beneficial because of their high T g values, frequently good stability is not obtained when these compounds are used as the sole stabilizers. It has been suggested that large polymers do not inhibit protein unfolding during dehydration because of steric hindrance effects that prevent effective hydrogen bonding to the protein in dried form. 42,56 With trehalose or stachyose formulations, drying method had little impact on secondary structure in the dried solid (Tab. 4), even though the surfactant-free spray dried preparation showed much more in-process aggregation than the other drying methods (Tab. 4). This result suggests that the formation of aggregates during a drying process and protein secondary structure in the final dried product are not necessarily strongly linked. Additionally, we note that the solid state FTIR technique is not sufficiently sensitive to pick up moderate levels of aggregated solid. 44 Drying method impacted protein secondary structure only with higher molecular weight stabi- DOI /jps

13 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE 175 Table 4. In-Process Stability of Met-hGH (Structural Correlation Coefficient Values Calculated from FTIR Studies, and Initial Aggregate Levels after Drying) and Rate of Physical Aggregation from Storage Stability Studies at 50 and 608C Formulation Treatment Spectral Correlation Coefficient (r) Initial Levels of Aggregates (%) SD Rate of Aggregation (k) a (%P/month 0.5 ) SE 1:3 hgh/trehalose Freeze dried <0.35 b Spray drying <0.35 b 1:3 hgh/trehalose Freeze dried (with Tween-20) Annealed/ freeze dried Spray dried Film dried <0.35 b 1:3 hgh/stachyose Freeze dried (with Tween-20) Annealed/ freeze dried Spray dried Film dried <0.35 b 1:3 hgh/dextran 40 Freeze dried Spray drying :3 hgh/dextran 40 Freeze dried (with Tween-20) Spray drying :3 hgh/ficoll 70 Freeze dried :3 hgh/ficoll 70 (with Tween-20) 508C 608C Spray drying Freeze dried Spray dried Film dried SD, standard deviation for replicate samples; SE, standard error. Gray shading: not measured. a Rates of physical aggregation are based on three time points (0, 1, and 3 months). b k values were less than the limit of detection based on assay errors. lizers. In surfactant-free formulations of dextran 40 and Ficoll 70, spray dried preparations seemed to have slightly less perturbation in secondary structure than their freeze dried counterparts (Tab. 4), even though in-process aggregation was much greater with the spray drying process. These results suggest possible structural damage that did not lead to aggregation during freezing in these systems. Most important, the results emphasize that retention of secondary structure in the dry product is not necessarily a good measure of in-process stability. Addition of surfactant to these preparations offered further protection against loss of protein secondary structure during both drying processes. However, protein secondary structure in spray dried preparations still remained superior to that in freeze dried preparations. FTIR studies were not carried out on film dried preparations. In-Process Stability Drying method had a significant impact on inprocess stability of Met-hGH (Tab. 4). Regardless of the presence or absence of surfactant, greater in-process aggregation of Met-hGH was observed with spray dried formulations. As expected, surfactant-free spray dried formulations suffered the greatest in-process aggregation. The initial large loss with spray drying probably occurred during the atomization step when a large air water interface is created, causing a sensitive protein such as Met-hGH (being surface active) to DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008

14 176 ABDUL-FATTAH ET AL. adsorb at the interface and partially unfold. 50,57 59 In-process aggregation was minimized by the addition of surfactant, as is consistent with the literature, 50,57 60 suggesting that Tween-20 was effective in protecting Met-hGH against unfolding at the air water interface. Additionally, the presence of surfactant reduced initial aggregate formation during freeze drying, indicating that Tween-20 was somewhat effective in protecting Met-hGH during freezing. Annealing in the frozen state, which subjected Met-hGH to an ice-water interface for a prolonged period, did not cause further aggregation in Met-hGH. This observation is important because conventional wisdom suggests caution in using a freezing process that involves annealing when freeze drying a sensitive protein. The concept is that annealing exposes the protein to a highly concentrated solution, well above T 0 g, for an extended period of time, thereby creating the molecular mobility needed to support degradation processes such as aggregation. Clearly, this is not an issue with Met-hGH, even though this protein is very sensitive to aggregation. However, it should also be noted that freeze drying above the collapse temperature does not necessarily cause significant protein degradation. 21 Moreover, it has recently been pointed out that protein unfolding is extremely slow even above T 0 g, where annealing is carried out. 61 It is possible that even at temperatures above T 0 g, the viscosity is sufficiently high that aggregation is slow on the time scale of the annealing process. Obviously, the difference between the annealing temperature and T 0 g is critical because viscosity will decrease sharply as the temperature exceeds T 0 g and will eventually be sufficiently low to kinetically allow unfolding on the timescale of annealing. It seems, however, that moderate excursions above T 0 g are sufficient to allow growth of ice crystals (i.e., Ostwald Ripening) but not sufficient to allow significant protein unfolding. It is also important to note that film drying caused 1% higher in-process aggregation in stachyose and Ficoll 70-based formulations than did freeze drying, perhaps because of the long residence time in a concentrated system at temperatures very much higher than T g. Thus, perhaps the difference between the system temperature and T g is critical; just above T g does not necessarily mean aggregation, but much above T g might constitute a problem. The type of stabilizer impacted in-process aggregation in freeze dried preparations, as evidenced by higher in-process aggregation with JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 freeze dried dextran-based preparations relative to the freeze dried preparations using trehalose and Ficoll 70. Storage Stability (Physical Stability) and Solid State Properties Percent of aggregates was determined at 0, 1, and 3 months storage at 508C and at 608C. Aggregate formation appeared to be at least bi-phasic (Fig. 4A), but degradation kinetics were linear with square root of time (Fig. 4B). Degradation in amorphous pharmaceuticals is frequently described by square root of time or stretched exponential kinetics. 62,63 Multiple linear regression was done using a general linear model (GLM) to plot and fit the assay data with different exponents of time. F-values for each GLM model with different time exponent showed that the model with square root of time kinetics fit the data most appropriately, confirming square root of time dependence (Tab. 5). The time dependence of percent aggregates, %P, were therefore analyzed using the linear equation: 63 p %P ¼ P o þ k ffiffi t (3) where P o is the initial level of aggregation and k is the apparent rate constant for aggregation. Results are summarized in Table 4. Met-hGH had high initial chemical impurities, roughly 20%, regardless of the presence of surfactant and there was some variation with preparation method. Based on past experience, Met-hGH is more susceptible to met oxidation than natural sequence hgh, but this result was not expected and could pose a problem if the aggregation rate during storage were highly correlated with the initial purity. To insure such was not the case, chemical purity of a number of samples was assessed using a reverse phase (RP) isocratic HPLC method developed by Riggin et al. 64 initially and after 3 months of storage at 508C, and the correlation between aggregation rate during storage at 508C and the initial purity was examined. We found a poor correlation between physical aggregation rate and both initial and final chemical purities (i.e., after 3 months at 508C), with the correlation coefficients being and for the correlations between initial purity and purity at the end of 3 months, respectively. Thus, the variation in chemical purity of Met-hGH in different samples is not a DOI /jps

15 STABILITY OF MeT-hGH IN THE AMORPHOUS SOLID STATE 177 Figure 4. Plots for physical degradation of freeze dried formulations with surfactant at 608C. Symbols key: triangles ¼ Ficoll 70, squares ¼ stachyose, diamonds ¼ trehalose. (A) Aggregation versus time. (B) Aggregation versus square root of time. significant factor in the variation of its aggregation rate during storage. Effect of Drying Method In the surfactant-free trehalose formulation, the freeze dried and spray dried preparations were very stable with no quantifiable degradation even after 3 months of storage at 508C, regardless of protein surface accumulation. In the surfactantfree dextran formulations, spray dried and freeze dried preparations showed similar physical stability ( p > 0.1 from Bonferroni test). With the surfactant-free Ficoll 70 formulation, the physical stability of the spray dried preparation was inferior to the freeze dried preparation ( p < 0.1 from Bonferroni test), even though protein secondary structure was similar in both. Total protein surface accumulation, however, was higher with the spray dried preparation. In formulations with surfactant, the film dried preparations showed superior physical stability relative to all other preparations regardless of the stabilizer used, consistent with a very low SSA DOI /jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008