Anal. Chem. 1997, 69, 4184-4190
Formulation Development and Primary Degradation Pathways for Recombinant Human Nerve Growth Factor Marian Eng,† Victor Ling,† Jonathan A. Briggs,‡,§ Karen Souza,‡ Eleanor Canova-Davis,† Michael F. Powell,‡ and Linda R. De Young*,‡,⊥
Department of Medicinal and Analytical Chemistry and Department of Pharmaceutical Research and Development, Genentech, Inc., 460 Point San Bruno Boulevard, South San Francisco, California 94080
The chemical and physical stabilities of recombinant human nerve growth factor (NGF) in aqueous solution were investigated between 5 and 37 °C and at pH 4.25.8. NGF chemical stability decreased with a decrease in pH due to Asp60-Pro61 cleavage, with the stability being greater in acetate buffer than in succinate buffer at each pH investigated. Aggregation was a significant degradation pathway at 37 °C, with the aggregation rate being greatest in succinate buffer at pH 5.8. Quantitation of NGF degradation by cation-exchange chromatography was complicated by the rearrangement of the NGF monomer variants into various mixed dimers over time. Treatment with dilute acid brought the dimer distribution rapidly to equilibrium, allowing NGF degradation to be accurately quantitated. An acetate-buffered formulation at pH 5.5 was investigated in more detail. To assist in degradation product identification, NGF degradation was accelerated with base, hydrogen peroxide, and temperature. These degradation products were shown to coelute on RP-HPLC with the variants found when the protein was stored at -70, 5, and 25 °C. By electrospray mass spectrometry, peptide maps, and LC/MS, these degradation products were shown to be monooxidized (Met37) and dioxidized (Met37 and Met92) NGF, with Met37 being more labile, deamidated NGF (Asn45), and NGF with Asp93 isomerized to β-Asp93. NGF can be stored in pH 5.5 acetate buffer at 5 °C for 1.5 years with less than 10% conversion to these degradation products, with Asp93 isomerization being the primary degradation pathway. Nerve growth factor (NGF) is a neurotrophic factor required for the growth and survival of sympathetic and sensory neurons.1,2 NGF has also been shown to support the growth, differentiation, and maintenance of cholinergic neurons in the central nervous system.1,2 Potential clinical indications for recombinant human NGF include peripheral sensory neuropathy and Alzheimer’s disease. For example, the systemic administration of NGF has been shown to reduce the sensory neuropathy induced by
administration of cisplatin and paclitaxel to mice.3,4 Clinical trials are currently underway to determine whether NGF administered to humans will improve sensory function in diabetic neuropathy. The gene for NGF encodes a protein of 120 amino acids, which associates through noncovalent forces5 with a dimer dissociation constant of 10-13 M or less.6,7 The recombinant human NGF used in these studies has, primarily, 118 amino acids, due to enzymatic processing at the C-terminus during production.8 Smaller amounts of species with 120 and 117 amino acids are also present, complicating the evaluation of protein stability due to the formation of heterodimers between the 117, 118, and 120 variants. Common chemical degradation pathways for proteins are methionyl oxidation, deamidation of asparagine, and aspartic acid isomerization. NGF contains two methionyl residues which are potential sites of oxidation, a single Asp-Gly sequence which is susceptible to aspartyl isomerization, a single Asp-Pro sequence which is susceptible to acid-catalyzed cleavage, and several AsnGly and Asn-Ser sequences which are prone to deamidation.9,10 A screen of NGF chemical and physical stability between pH 4.2 and 5.8 at -70 and 37 °C was done to choose optimal solution conditions. A previous study had shown NGF stability to decrease above pH 6.0 (X. Lam and T. Nguyen, personal communication). The chemical degradation pathways for NGF were then elucidated for a pH 5.5 acetate formulation. EXPERIMENTAL SECTION Materials. Recombinant human nerve growth factor (NGF) was produced in Chinese hamster ovary cells and purified by reversed-phase high-performance liquid chromatography (RPHPLC) and ion-exchange chromatography (IEC) as described previously.8 The gene encodes for a 120-amino acid protein. However, proteolytic cleavage during cell culture results in conversion primarily to a 118-amino acid product. Smaller amounts of the 120- and 117-amino acid species copurified with
Department of Medicinal and Analytical Chemistry. Department of Pharmaceutical Research and Development. § Current address: Inhale Therapeutic Systems, 1060 East Meadow Circle, Palo Alto, CA 94303. ⊥ Current address: AeroGen, Inc., 1310 Orleans Dr., Sunnyvale, CA 94089. (1) Hefti, F. J. Neurobiol. 1994, 25, 1418-1435. (2) Levi-Montalcini, R. Science 1987, 237, 1154-1162.
(3) Apfel, S. C.; Lipton, R. B.; Arezzo, J. C.; Kessler, J. A. Neurology 1991, 28, 87-90. (4) Apfel, S. C.; Arezzo, J. C.; Lipson, L. A.; Kessler, J. A. Ann. Neurol. 1992, 31, 76-80. (5) McDonald, N. Q.; Lapatto, R.; Murray-Rust, J.; Gunning, J.; Wlodawer, A.; Blundell, T. L. Nature 1991, 354, 411-414. (6) Bothwell, M. A.; Shooter, E. M. J. Biol. Chem. 1977, 252, 8532-8536. (7) Timm, D. E.; de Haseth, P. L.; Neet, K. E. Biochemistry 1994, 33, 46674676. (8) Schmelzer, C. H.; Burton, L. E.; Chan, W.-P.; Martin, E.; Gorman, C.; CanovaDavis, E.; Ling, V. T.; Sliwkowski, M. B.; McCray, G.; Briggs, J. A.; Nguyen, T. H.; Polastri, G. J. Neurochem. 1992, 59, 1675-1683. (9) Geiger, R.; Clarke, S. J. Biol. Chem. 1987, 262, 785-794. (10) Manning, M. C.; Patel, K.; Borchardt, R. T. Pharm. Res. 1989, 6, 903-918.
4184 Analytical Chemistry, Vol. 69, No. 20, October 15, 1997
S0003-2700(97)00401-0 CCC: $14.00
† ‡
© 1997 American Chemical Society
the 118-amino acid homodimer. HPLC grade acetonitrile and trifluoroacetic acid (TFA) were used for RP-HPLC. Endoproteinase lysine-C was purchased from Wako Pure Chemical Industries. All other chemicals were USP grade. Methods. NGF was dialyzed into 10 mM sodium acetate, 142 mM sodium chloride, at pH 5.0 and 5.8, or 10 mM sodium succinate, 142 mM NaCl, at pH 4.2, 5.0, and 5.8, and adjusted to 10 mg/mL. Surfactant was also added to a succinate pH 5.0 solution (10 mM sodium succinate, 142 mM NaCl, 0.05% Tween 20) to ascertain any effects on NGF aggregation. NGF solutions studied at pH 5.5 contained 2 mg/mL NGF in 10 mM sodium acetate, 142 mM NaCl. Vials were aseptically filled with 0.3 mL of the 10 mg/mL NGF solution and incubated at 5, 25, and 37 °C. Controls were stored at -70 °C as no significant degradation had been observed at this temperature or during the freeze-thaw process. For studies at pH 5.5, vials were filled with 0.7 mL of 2 mg/mL NGF and incubated at -70, 5, 25, and 37 °C. At various time points, samples were transferred to -70 °C and stored until analysis. HPLC Analysis for Formulation Screening. Cation-exchange HPLC was carried out on a HP 1090 system using a Tosohas sulfopropyl TSK-SP-5PW (7.5 mm × 75 mm) column with 10-mm particles. Mobile phases were (A) 10 mM sodium phosphate, 5% (v/v) acetonitrile, pH 7.0, and (B) A + 1.0 M ammonium chloride. NGF was eluted at 35 °C (0.5 mL/min) with a linear gradient of 20-40% B from 5 to 60 min. The control and 1.6-year samples at 5 °C were also assayed after “acid-treatment” to bring the distribution of monomer variants in the dimers to equilibrium. These samples were adjusted to pH 3.5 with HCl and incubated at 37 °C for 2 h (results at 2 and 4 h were equivalent). A YMC C4 (4.6 mm × 250 mm, 5 µm particles with 300-Å pores) column was used for RP-HPLC on a HP 1090 system at 25 °C. NGF was eluted (0.5 mL/min) using a linear gradient of 26-30% B (A ) 0.05% TFA in water and B ) 0.05% TFA in acetonitrile) from 5 to 40 min. Size-exclusion HPLC (SEC) was performed using a Perkin Elmer Series 410 Bio LC pump with a Perkin Elmer LC 90 spectrophotometric UV detector and a Tosohas TSK 2000 SWXL (7.8 mm × 300 mm, 5 µm) column. This SEC column was eluted at 0.5 mL/min using a 0.2 M potassium phosphate, 0.45 M potassium chloride mobile phase at pH 7.0. For SEC, UV detection was at 280 nm, and for RP-HPLC and IEC, detection was at 214 nm. For all assays, 30 µg of NGF was injected. Neurite Outgrowth Assay. The biological activity of NGF was determined using the PC12 assay developed by Greene11 and modified as described by Schmelzer et al.8 Hemolysis. All formulations were tested for hemolytic activity. The hemolysis procedure was that of Reed and Yalkowsky,12 except that equal volumes of washed human red blood cells and formulation were incubated at 37 °C for 30 min before analysis. Degradation Product Analysis. Preparation of Degraded NGF. To chemically oxidize NGF, 1.1 µL of 3% hydrogen peroxide was added to 55 µL of 2 mg/mL NGF in the pH 5.5 acetate formulation. This sample was incubated at ambient temperature for 2 h before preparative fractionation and analysis. NGF was thermally degraded by incubating at 25 °C for 5.6 months in pH 5.5 acetate buffer. A deamidated NGF variant was generated by incubating a NGF sample at pH 8.1 for 16-18 h.
Preparative and Analytical HPLC. Degraded NGF was fractionated and collected using a Synchropak C4 (6.5 µm, 300-Å pore, 0.46 cm × 25 cm) preparative RP-HPLC column run at 0.5 mL/ min and 25 °C on a HP1090 system. The degraded NGF (100 µg) was loaded onto the column and eluted with a 15-min isocratic wash of 27.5% B (A ) 0.05% TFA in water, B ) 0.05% TFA in acetonitrile), followed by a linear gradient of B at 0.76%/min for 5 min, 1.9%/min for 5 min, and then 8%/min for 5 min. Fractions were collected and pooled from consecutive runs to obtain sufficient material for analysis. The fractions were evacuated to dryness in a SpeedVac and reconstituted with water to a concentration of approximately 0.5 mg/mL and stored at -70 °C. Analytical RP-HPLC was run on a YMC C4 (5 µm, 300 Å, 0.46 cm × 25 cm) column using the preparative RP-HPLC method. UV detection for both methods was at 214 nm. Mass Spectrometry. Samples were prepared by exchanging 5 µg of each collected fraction into deionized water by repeated washings in a Centricon 10 (Amicon), followed by evaporation to dryness in a SpeedVac. The samples were then reconstituted with 30% acetonitrile in water with 0.5% acetic acid to a final protein concentration of approximately 30 pmol/µL. The solutions were infused at 3 µL/min into the ionspray source of a Sciex API III triple-quadrupole mass spectrometer. Masses were calibrated using a mixture of poly(propylene glycol)s. The mass spectra were analyzed using Reconstruct and Hypermass software (Sciex) to calculate molecular masses from multiply charged ions. Endoproteinase Lysine-C Peptide Maps. A 20-µg aliquot of each NGF fraction was dried and reconstituted with 20 µL of 100 mM Tris buffer, pH 8.2. Endoproteinase lysine-C was added at an enzyme-to-substrate ratio of 7.5% (w/w) at t ) 0 and 2 h during the 37 °C incubation. After 5 h of incubation, the digestion was terminated by freezing. The digested protein (2 µg) was loaded onto a Zorbax C8 (5 µm, 300 Å, 2.1 mm × 150 mm) reversedphase column and eluted at 0.2 mL/min at 40 °C using the following gradient: 0.1% TFA in water for 10 min, followed by a linear gradient of 0.1% TFA in acetonitrile of 0.3%/min for 10 min, 2.0%/min for 10 min, 0.2%/min for 10 min, 0.5%/min for 4 min, 1.0%/min for 3 min, 2%/min for 5 min, and 5.0%/min for 4 min. UV detection was at 214 nm. Peptide masses were analyzed online using a postcolumn splitter to divert effluent to a Sciex API III triple-quadrupole mass spectrometer. N-Terminal Sequence Analysis. The modified L8 peptide (100 pmol) from the NGF sample degraded at 25 °C was subjected to seven cycles of Edman degradation in an Applied Biosystems Model 477A protein sequencer. The PTH amino acid derivatives were analyzed on-line using the Applied Biosystems Model 120A system. In addition, NGF in 10 mM succinate buffer at pH 4.2 was incubated at 37 °C for 3.5 months, run on reduced SDSPAGE (100 mM), blotted onto PVDF, and analyzed as above.
(11) Greene, L. A. Brain Res. 1977, 133, 350-353. (12) Reed, K.; Yalkowsky, S. J. Parenter. Sci. Technol. 1987, 41, 37-39.
(13) Timm, D. E.; Neet, K. E. Protein Sci. 1992, 1, 236-244. (14) Moore, J. B.; Shooter, E. M. Neurobiology 1975, 5, 369-381.
RESULTS AND DISCUSSION Temperature, Buffer, and pH Dependence of NGF Chemical and Physical Stability. Aggregation of NGF. The dimer/ monomer equilibrium constant for murine NGF is less than 10-13 M at pH 4-7.6,7,13,14 Recombinant human NGF, with a similar equilibrium constant,7 remains as a dimer (retention time, 16.0 min) in the neutral pH SEC assay. A small amount of aggregated NGF (tetramer based on molecular weight standards) was
Analytical Chemistry, Vol. 69, No. 20, October 15, 1997
4185
Figure 1. Dependence of NGF aggregate formation at 37 °C on formulation buffer and pH, quantitated by size-exclusion chromatography, (]) succinate pH 4.2; (4) succinate pH 5.0; (×) succinate pH 5.0 with 0.05% Tween 20; (0) succinate pH 5.8; (2) acetate pH 5.0; and (9) acetate pH 5.8.
observed in the control sample (retention time, 14.5 min), and the tetramer peak area increased with time at 37 °C. A leading shoulder on this peak (retention time, 13.5 min), indicating larger aggregates, was observed for all formulations after 38 days at 37 °C (data not shown). The time dependencies of aggregate formation for the various formulations are shown in Figure 1. There was no significant difference in the aggregate formation rates except for the succinate pH 5.8 formulation, which had the greatest aggregation rate. The surfactant Tween 20 offered no protection against aggregation. NGF Monomer and Degradation Product Quantitation. The NGF used in this formulation screening study consisted of a 1:9:1 ratio of three monomeric polypeptides containing 120, 118, and 117 amino acids. At pH 5.0, the 117 variant has two fewer positive charges (cleavage of Arg118-Arg119-Ala120) and the 118 variant one fewer positive charge (cleavage of Arg119-Ala120) than the 120 parent. NGF exists as a monomer in the acidic RP-HPLC mobile phase.8 The RP-HPLC elution order was 120 < 118 < 117. RPHPLC chromatograms for NGF stored in pH 5.0 succinate buffer for 38 days at -70 and 37 °C are shown in Figure 2. The loss of total NGF peak area (120 + 118 monomers) with incubation time at 37 °C can be described as a first-order process. The apparent first-order rate constants for this loss of peak area are tabulated in Table 1. The 117 peak area was not included in the NGF total peak area due to coelution of degradation products with this peak at elevated temperatures. As shown in Table 1, NGF stability decreased as the pH was lowered, with the stability being greatest in acetate buffer. Tween 20, added to potentially decrease NGF aggregation, did not affect NGF stability in succinate buffer at pH 5.0. NGF has a potentially acid-labile Asp60-Pro61 linkage in the cystine-knot loop. A different preparation of NGF (with less than 5% Ser1-to-Gly1 conversion, see below) at pH 4.2 was incubated at 37 °C for 3.5 months, assayed by reduced SDS-PAGE, and blotted onto PVDF for N-terminal amino acid sequence analysis. Three low molecular weight bands showed equal amounts of Ser1 and Pro61 as the N-terminal amino acid. This confirmed that a significant degradation pathway for NGF at pH 4.2 was Asp60-Pro61 cleavage. At pH 5.5, this was not a significant degradation pathway. 4186
Analytical Chemistry, Vol. 69, No. 20, October 15, 1997
Figure 2. Representative RP-HPLC spectra for NGF in succinate buffer at pH 5.0, (a) -70 °C control and (b) after 38 days of incubation at 37 °C. Table 1. Apparent First-Order Rate Constants for NGF Degradation at 37 °C As Determined by RP-HPLC buffer succinate succinate + Tween 20 acetate
pH
k × 10-3 (day-1)
4.2 5.0 5.0 5.8 5.0 5.8
22 ( 1 11 ( 0.6 11 ( 0.7 5.7 ( 1 7.9 ( 0.8 4.0 (0.3
NGF Dimer Distribution. The 118/118 and 117/120 NGF dimers have the same net charge and, therefore, coelute on IEC during NGF purification. This results in an initial nonequilibrium distribution of the monomer variants in NGF dimers in the NGF product. The 117/120 and 118/118 dimers can dissociate to 117, 118, and 120 monomers and randomly reassociate to the 118/ 118 dimer, with smaller amounts of the 117/118, 118/120 and 117/ 117, and 120/120 dimers.8,14 Due to the different charges on the monomers, the expected elution order of these dimers on cationexchange chromatography is
117/117 < 117/118 < 118/118 ) 117/120 < 118/120 < 120/120
Representative IEC chromatograms for NGF at pH 5.0 in succinate buffer after 38 days at -70 and 37 °C are shown in Figure 3A. The most populated dimers are distinguishable.8 NGF undergoes an N-terminal modification during fermentation, where Ser1 is converted to Gly1,15 resulting in delayed elution on IEC (Figure 3A). NGF was quantitated here as the sum of the 118/ 118 homodimer and the 118/118 dimer with a Ser-to-Gly conversion in a single chain of the dimer (and any coeluting 117/120 variants). The 117/118 and 118/120 peak areas were not included due to degradation products coeluting with the 117/118 peaks (data not shown). The rate of loss of NGF at 37 °C, as monitored (15) Canova-Davis, E.; Ling, V.; Eng, M.; Skieresz, S. Peptides: Chemistry, Structure and Biology ; Escom Science Publishers: Leiden, The Netherlands, 1993; pp 230-231.
Figure 4. Comparison of NGF (s) -70 °C control and (- - -) 5 °C IEC spectra after 1.6 years of incubation in acetate buffer at pH 5.0, with (A) no acid treatment and (B) acid treatment of samples prior to analysis.
Figure 3. (A) Representative IEC spectra for NGF in acetate buffer at pH 5.0 after 38 days of incubation at (s) -70 and (- - -) 37 °C. Each dimer appears as a triplet in the chromatogram due to N-terminal Ser-to-Gly (S1G) conversion.15 The earliest peak in the triplet is the parent dimer, followed by a dimer with a single Ser-to-Gly conversion, and finally a dimer with a Ser-to-Gly conversion in both chains. (B) Time dependence of the loss of NGF 118/118 and 117/120 dimers upon incubation at 37 °C, (4) succinate pH 5.0; (×) succinate pH 5.0 with 0.05% Tween 20; (0) succinate pH 5.8; (2) acetate pH 5.0; and (9) acetate pH 5.8.
by IEC, is shown in Figure 3B. The degradation kinetics were multiphasic. The loss in main peak area before 13 days was largely due to rearrangement of the monomer variants between the possible dimer types. A detailed study of other NGF degradation pathways observed by IEC for loss of NGF after 13 days has not been done. By IEC, NGF was most stable in the acetate formulations at pH 5.0 and 5.8, which had similar stability. The accelerated stability data obtained at 37 °C showed that NGF was most stable in acetate buffer. The protein stability was also investigated at 5 °C. After 1.6 years of storage at 5 °C, the apparent first-order rate constants for NGF degradation, as determined by RP-HPLC, were 1.4 × 10-4 ( 1.7 × 10-5 and 6.2 × 10-5 ( 1.1 × 10-5 day-1 at pH 5.0 and 5.8, respectively. IEC showed that NGF stability was approximately the same at pH 5.0 and 5.8, consistent with the 37 °C IEC data. Aggregation of the NGF dimers was not a significant degradation pathway at 5 °C, with only a 1% increase in aggregation over 1.6 years of storage
at 5 °C. For all methods, the rate of degradation of NGF was slower at 5 than at 37 °C. The interpretation of the IEC data at both 5 and 37 °C was complicated by dimer exchange, the exchange rate being slower at the lower temperature. To improve IEC quantitation, the dimer distribution was brought to equilibrium by incubation at pH 3.5 for 2 h at 37 °C prior to IEC analysis.8,14,16 No new peaks in the IEC chromatograms were observed after this treatment. The acetate pH 5.8 samples after 1.6 years of incubation at 5 °C are compared with controls before and after “acid treatment” in Figure 4. Quantitation after acid treatment showed that 94 and 92% NGF remains after 1.6 years at 5 °C at pH 5.0 and 5.8, respectively, compared to 84 and 87% without acid treatment. For comparison, RP-HPLC analysis showed 93 and 96% NGF remaining at pH 5.0 and 5.8, respectively. An alternative method of data analysis, where the peak areas for the 117/118, 118/118, and 118/120 dimers are summed, may not be used due to coelution of degradation products with the 117/118 dimer. Bioactivity and Hemolysis Testing. The hemolytic activity of each of the NGF formulations was tested. None of the formulations showed significant red blood cell hemolysis (
