Effects of pH, temperature, and sucrose on benzyl alcohol-induced aggregation of recombinant human granulocyte colony stimulating factor

Authors

  • Renuka Thirumangalathu,

    1. Department of Pharmaceutical Sciences, Center for Pharmaceutical Biotechnology, University of Colorado Health Sciences Center, University of Colorado, Denver, Colorado 80262
    Search for more papers by this author
  • Sampathkumar Krishnan,

    1. Department of Pharmaceutics, Amgen, Inc., Thousand Oaks, California 91320
    Search for more papers by this author
  • David N. Brems,

    1. Department of Pharmaceutics, Amgen, Inc., Thousand Oaks, California 91320
    Search for more papers by this author
  • Theodore W. Randolph,

    1. Department of Chemical and Biological Engineering, University of Colorado, Boulder, Colorado 80309
    Search for more papers by this author
  • John F. Carpenter

    Corresponding author
    1. Department of Pharmaceutical Sciences, Center for Pharmaceutical Biotechnology, University of Colorado Health Sciences Center, University of Colorado, Denver, Colorado 80262
    • Department of Pharmaceutical Sciences, Center for Pharmaceutical Biotechnology, University of Colorado Health Sciences Center, University of Colorado, Denver, Colorado 80262. Telephone: 303-315-6075; Fax: 303-315-6281
    Search for more papers by this author

Abstract

Antimicrobial preservatives (e.g., benzyl alcohol), which are required in multidose formulations, can induce protein aggregation. In this study, the mechanism of benzyl alcohol-induced aggregation of recombinant human granulocyte colony-stimulating factor (rhGCSF) was investigated by determining the effects of temperature, pH, and sucrose on this process. rhGCSF was incubated at 25 and 37°C and at pH 7.0 (phosphate-buffered saline, PBS) and pH 3.5 (HCl). Benzyl alcohol (0.9% w/v) accelerated aggregation of rhGCSF at pH 7.0, an effect that was much greater at 37°C than at 25°C and partially counteracted by 1.0 M sucrose. At pH 3.5, benzyl alcohol did not induce aggregation of rhGCSF. Spectroscopic studies showed that 0.9% benzyl alcohol altered the tertiary structure of rhGCSF at both pH, without detectably altering secondary structure. Structural perturbation was greater at 37°C than at 25°C. At both pH 7.0 and 3.5, the hydrogen-deuterium (H–D) exchange rate for rhGCSF was increased by 0.9% benzyl alcohol. Sucrose (1.0 M) partially counteracted the benzyl alcohol-induced perturbation of tertiary structure and the increase in H–D exchange rate. Thus, benzyl alcohol accelerates aggregation of rhGCSF at pH 7.0, because it favors partially unfolded aggregation-prone conformations of the protein. Sucrose partially counteracts benzyl alcohol-induced rhGCSF aggregation by shifting the molecular population away from these species and towards more compact conformations. We postulate that the absence of aggregation at pH 3.5, even with benzyl alcohol-induced structural perturbation, is due to the unfavorable energetics of intermolecular interactions (i.e., colloidal stability) between rhGCSF molecules at this pH. © 2006 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 95: 1480–1497, 2006

Abbreviations:

rhGCSF, recombinant human granulocyte colony-stimulating factor; H–D, hydrogen-deuterium; PBS, phosphate-buffered saline; SE-HPLC, size exclusion high performance liquid chromatography; CD, circular dichroism; UV, ultra violet; IR, infrared.

INTRODUCTION

As is the case with other classes of drugs, for therapeutic protein products it is important to maximize patient convenience and compliance, and to minimize the costs of therapy. An effective strategy to meet these goals is to create multidose formulations, which allow for removal of several doses of drug from the same vial. As a result, dose sizes can be adjusted as appropriate for a given patient. Furthermore, because essentially all of the product in a given vial can be utilized, product waste is minimized, leading to important cost savings. Antimicrobial preservatives such as benzyl alcohol must be included in multidose formulations of therapeutic proteins to inhibit microbial growth.1 Preservatives are also required for certain protein drug delivery systems, for example, injection pens that are used for multiple doses2 and mini pumps that are used for continuous infusion.3 However, preservatives can greatly accelerate formation of nonnative protein aggregates.4–8 In addition to the potential loss of biological activity due to aggregation, the presence of aggregates can cause adverse responses in patients.9,10 A successful multidose therapeutic protein formulation must both maintain product sterility and minimize the destabilizing effects of the preservative on the protein. Therefore, it is vital to understand the mechanism for preservative-induced protein aggregation in order to develop a rationale for choosing effective stabilizers (e.g., sucrose) and solution conditions (e.g., pH) to inhibit the process.

Earliest studies on preservative-induced protein aggregation documented that preservatives caused a reduction in thermal stability of proteins4,7,11 and perturbation of protein tertiary structure.5 However, it was not determined if the conformational changes observed in the presence benzyl alcohol occurred before or after aggregation. More recently, Tobler et al.,12 reported with mass spectrometric analysis of interferon-γ that benzyl alcohol accelerated the rate of H–D exchange for the protein. Consistent with this observation, we have proposed and confirmed—with a model β-sheet protein (recombinant human interleukin-1 receptor antagonist)—the hypothesis that benzyl alcohol interacts hydrophobically with partially unfolded protein molecules from which protein aggregates are formed.8 Thus, according to the Wyman Linkage function, preferential interaction of the preservative with partially unfolded molecules results in a shift in the population of protein molecules towards these aggregation-prone species and, hence, an increase in aggregation rate. In further support of this mechanism, we documented that sucrose, which favors the most compact conformations in the native state ensemble, could counteract preservative-induced protein aggregation.8

The purposes of the current study are twofold. First, we tested this hypothesis using a model α-helical protein, recombinant human granulocyte colony-stimulating factor (rhGCSF). This protein is from the four-helix bundle family of proteins, which includes many growth factors and cytokines.13 Second, we wanted to use investigations of the effects of temperature, pH, and sucrose to gain further insights into the mechanism for preservative-induced protein aggregation. Hydrophobic interactions, which are generally strengthened at increased temperatures,14 appear to govern benzyl alcohol interaction with protein molecules.8 Thus, we hypothesize that aggregation of rhGCSF will be enhanced at elevated temperature by favoring a greater benzyl alcohol-induced increase in partially unfolded molecules than that arising at lower temperature.

For many proteins, pH is an important solution condition governing rates of aggregation.15 pH can affect the thermodynamic stability of the native state and the protein's conformation, as well as the energetics of intermolecular interactions. With rhGCSF we have previously documented that the protein aggregated rapidly in aqueous solution at pH 7, but over the same time frame did not aggregate at pH 3.5.16,17 The protein's free energy of unfolding and conformation were not significantly different at the two pHs. Instead, charge–charge repulsion (colloidal stability) at pH 3.5 accounted for inhibition of aggregation. In contrast, at pH 7.0 intermolecular interactions of rhGCSF are attractive, favoring aggregation. For the current study, we propose that preservative-induced aggregation similarly will be modulated by solution pH. Thus, for an increase in partially unfolded protein molecules caused by benzyl alcohol will result in acceleration of rhGCSF aggregation at pH 7, whereas at pH 3.5, charge–charge repulsion between these molecules will inhibit aggregation.

Finally, we have previously documented with interleukin-1 receptor antagonist that the thermodynamic stabilizer sucrose can counteract the structural perturbations and aggregation induced by benzyl alcohol. Sucrose has also been found to be effective in reducing the aggregation rate of proteins in the presence of compounds such as urea by counteracting the effects of the denaturant.18 We have shown previously that sucrose inhibits rhGCSF aggregation arising under physiological conditions (pH 7, phosphate-buffered saline) by favoring the more compact species in the protein's molecular population.16 Therefore, we hypothesize that sucrose should at least partially counteract the benzyl alcohol-induced increase in levels of partially unfolded molecules and the resulting acceleration of aggregation.

The depletion of monomeric rhGCSF during incubation was quantified using size exclusion high performance liquid chromatography (SE-HPLC). The effects of temperature and solution conditions on secondary structure of the protein prior to aggregation, and the structural transitions occurring during aggregation, were characterized using infrared (IR) spectroscopy. Near ultra violet (UV) circular dichroism (CD) was used to monitor the tertiary structural changes in rhGCSF prior to protein aggregation. Preservative induced shifts in molecular population towards partially unfolded molecules were investigated by determining H–D exchange rates for rhGCSF in real-time with IR spectroscopy.

MATERIALS AND METHODS

Materials

Pharmaceutical quality rhGCSF (purity >99%) was a generous gift from Amgen, Inc., (Thousand Oaks, CA). The protein was obtained as a stock solution of 3.0 mg/mL rhGCSF in 10 mM sodium acetate and 100 mM sodium chloride (NaCl) at pH 3.6 and stored at 4°C. Sodium phosphate, deuterium oxide (99.0% D), deuterium chloride (35% wt in D2O), N-acetyl-tryptophanamide and N-acetyl-tyrosinamide were purchased from Sigma chemical company (St. Louis, MO). Benzyl alcohol was purchased from Fisher Scientific (Pittsburg, PA) and high purity sucrose was obtained from Pfanstiehl Laboratories (Waukegan, IL). All chemicals were of reagent grade or higher quality.

Aggregation Studies

A 4.0 M stock solution of sucrose (pH 7.0) and a 5% w/v stock solution of benzyl alcohol were prepared in 10 mM sodium phosphate and 150 mM NaCl at pH 7.0 (PBS). A 5% w/v stock solution of benzyl alcohol was prepared separately in pH 3.5 (HCl). rhGCSF stock solution was dialyzed separately against excess of PBS. rhGCSF (1.5 mg/mL) was incubated at 25 and 37°C without agitation as: rhGCSF in buffer alone, rhGCSF and 1.0 M sucrose, rhGCSF and 0.9% (w/v) benzyl alcohol, rhGCSF with 1.0 M sucrose, and 0.9% (w/v) benzyl alcohol. In addition, the stock solution was dialyzed into pH 3.5 solution (HCl) and diluted to 1.5 mg/mL. The solution was incubated alone and in combination with 0.9% benzyl alcohol without agitation at 25 and 37°C. For each time point, triplicate 100 µL samples were incubated in 250 µL polypropylene eppendorf tubes with separate tubes for each time point. Samples at each time were centrifuged at 13000g for 10 min (4°C) to separate the soluble protein from the insoluble precipitate. The supernatant was analyzed for the level of monomeric protein using SE-HPLC (HP-1090) by injecting 25.0 µL of the supernatant onto a TSK-GEL G2000SWxL column. The mobile phase consisted of 10 mM sodium phosphate and 150 mM NaCl with a flow rate of 0.6 mL/min. UV detection of the eluate was done at 280 nm. The level of soluble protein remaining in the supernatant for each time of incubation was quantitated as previously described.8,16

Secondary Structural Analysis

Secondary structure of rhGCSF was monitored using IR spectroscopy, in the conformationally sensitive amide I region, at 25 and 37°C. The protein stock solution obtained after dialysis in PBS or pH 3.5 HCl was concentrated to approximately 45.0 mg/mL using a 10 kDa Centricon concentration device (Millipore). Such high initial concentration is required to allow for dilution to a desired final protein concentration when combined with other excipients during sample preparation. Real-time IR measurements were performed for rhGCSF (concentration = 15 mg/mL) at pH's 3.5 and 7 for various formulations for 2 h at 25 and 37°C. For each spectrum, a 64-scan interferogram was obtained in the single beam mode. The sample cell consisted of CaF2 windows with a fixed path length of 6.0 µm. The corresponding spectrum for the buffer was obtained under identical conditions. The temperature was controlled with a Peltier temperature controller unit, and sample temperature was measured directly using a thermocouple inserted into a depression in the CaF2 window. For analysis, the spectra were initially converted to absorbance signals followed by subtraction of the corresponding buffer spectrum and processed as previously described.19,20 Far-UV CD spectra could not be used in the secondary structural examination of the protein due to the strong absorbance of benzyl alcohol in the far UV region.6

Tertiary Structural Analysis

Tertiary structure of rhGCSF was examined using second derivative UV, fluorescence and near UV CD spectroscopies under the same solution and formulation conditions used for the IR studies. Near UV CD spectra (270–340 nm) for the protein were obtained using an Aviv (Model 62DS) spectrometer equipped with a Peltier temperature control unit. The spectra were collected in a 1 cm path length cuvette using a protein concentration of 1.0 mg/mL at 25 and 37°C. Each individual scan was collected every 0.5 nm with an averaging time of 5 s and 1.0 nm bandwidth. The data obtained after the subtraction of the appropriate buffer spectrum was expressed as mean residue ellipticity, [θ], using a mean residue weight of 107.21 To improve resolution and signal quality each spectrum was the average of three separate scans. Fluorescence spectra for the protein were acquired using Aviv (Model ATF-105) fluorescence spectrophotometer at 25 and 37°C in a 1.0 cm quartz fluorescence cells using an rhGCSF concentration of 0.1 mg/mL. The protein was excited at 280 nm and the emission spectra were collected between 300–450 nm. The excitation and emission bandwidth was set to 4.0 and 8.0 nm, respectively. UV absorbance scans (from 200 to 500 nm) were collected in an Agilent 8453 UV–visible spectrometer (Palo Alto, CA), equipped with a Peltier temperature control unit. All measurements were made using rhGCSF concentration of 1.0 mg/mL in a 1.0-cm path length cuvette at 25 and 37°C and under both pH conditions. The data were processed using Grams software. All data were initially truncated between 250 and 320 nm. Subsequent to this step, the data was interpolated and the second derivative spectrum was calculated using Savitzky–Golay function. All spectra provided are the average of three separate scans.

Both temperature and solvent polarity can have potential effects on the UV absorbance and the fluorescence signal. Additives such as benzyl alcohol have substantial effects on solvent polarity, which may influence protein spectra. In order to validate if results from second derivative absorbance and fluorescence spectroscopies are truly representative of the effects of benzyl alcohol on the protein's conformation only, control samples were run under identical conditions. The control sample consisted of N-acetyl tryptophanamide and N-acetyl-tyrosinamide in equimolar ratios as their composition in rhGCSF (2:3, tryptophan:tyrosine) and at both pHs and temperature conditions. The UV absorbance and fluorescence scans were collected using the same parameters used for the protein sample and processed identically.

Hydrogen-Deuterium (H–D) Exchange Studies

H–D exchange was determined in real-time with a Bomem Prota IR spectrometer. Deuterated buffer and stock sucrose solutions were prepared as previously described.16 A 45-mg/mL stock solution of rhGCSF in H2O-based PBS or pH 3.5 HCl was diluted with the deuterated buffer (pD 6.6 or pD 3.1) to achieve a final ratio of 30% rhGCSF (at approximately 15.0 mg/mL) and 70% deuterated buffer by volume. The final solution was placed immediately in a sample cell composed of CaF2 windows separated with 12.0 µm polypropylene spacer. The time frame between sample preparation and first spectral acquisition was about 80 s. All spectra were acquired every 5 min initially for the first 40 min and then for every 20 min. Samples were incubated at 25 or 37°C, and H–D exchange was measured for 2 h, after which sample evaporation precluded acquisition of spectra. Attempts to measure the H–D exchange rate for the protein in combination with benzyl alcohol at 37°C were complicated by the severe aggregation of the protein even as early as in the first scan.

For each spectrum, 64 scans were collected using the IR spectrophotometer in the single beam collection mode. All spectra were initially converted to absorbance signals. After subtraction of respective buffer absorbance, the spectra were truncated between 1800 and 1350/cm and area normalized for subsequent analysis.22,23 The extent of H–D exchange reaction was determined by following the amide II/amide I ratio, tracking a loss of NH stretching according to a previously established criteria.20,22

Thermal Stability Using UV–Visible Spectroscopy

The effects of pH and benzyl alcohol on the thermal stability of rhGCSF (0.25 mg/mL) were determined using an Agilent 8453-UV–visible spectrophotometer. Scans were collected against a solvent matched reference over the temperature range 14–86°C at every 2°C increment. Samples were allowed to equilibrate for 2.0 min at each temperature prior to spectral acquisition. The absorbance spectra were processed to calculate the second derivatives as previously described.8 The second derivative peak shows two maxima centered on 287 and 295 nm and two minima at 283 and 290.5 nm due to contribution from tryptophan/tyrosine and tryptophan, respectively.

Thermostability of rhGCSF was monitored by following the change in tryptophan peak position as a function of temperature. Aggregation of the protein was monitored by the increase in optical density at 350 nm. To investigate the intrinsic effect of temperature on spectral peaks, control samples containing 3:2 ratios of N-acetyl tyrosinamide and N-acetyl tryptophanamide were run under identical conditions and the tryptophan minima at ∼290 nm was monitored as a function of temperature.

RESULTS

Aggregation Studies

The levels of monomeric rhGCSF remaining in solution during incubation at pHs 3.5 and 7 for experiments conducted at 25 and 37°C, respectively, are shown in Figure 1A and B. We have previously reported on a study of the aggregation of rhGCSF under physiological conditions (pH 6.9 in PBS),16 in which the extent of aggregation during incubation at 37°C for the protein in buffer alone was much greater than that seen in the present study (Fig. 1B). We hypothesize that this discrepancy can be attributed to differences in sample preparation and handling. In the present study, samples were prepared and incubated in separate Eppendorf tubes for each time point (Materials and Methods). In the earlier study, for each sample replicate a single tube was used for the entire incubation. At each time point, after resuspension of the contents in the Eppendorf tube by agitation, an aliquot was withdrawn for analysis. Repeated sample agitation and sample withdrawal from a single tube may accelerate aggregation of rhGCSF during incubation. To test this hypothesis, we replicated the method of the earlier study and incubated rhGCSF solution in a single eppendorf tube per sample replicate for the 5-day study (Method 1) experiment. For comparison, we employed the protocol of the current study, in which sample replicates are placed into separate Eppendorf tubes for each time point (Method 2). Samples were incubated at pH 3.5 and 7.0 at 37°C (Fig. 2). A much more rapid loss of monomer due to aggregation was noted with Method 1 compared to Method 2. This result clearly emphasizes that sample preparation and handling can have a profound effect on aggregation of rhGCSF during incubation in aqueous solution.

Figure 1.

Loss of native rhGCSF due to aggregation followed as a function of incubation time by size-exclusion chromatographic analysis. (A) At 25°C. (B) At 37°C. Legends: circles, rhGCSF alone (pH 7.0); upward triangle, rhGCSF with 1 M sucrose (pH 7.0); downward triangle, rhGCSF with 0.9% benzyl alcohol (pH 7.0); squares, rhGCSF with 0.9% benzyl alcohol and 1 M sucrose (pH 7.0); diamonds, rhGCSF alone (pH 3.5); hexagons, rhGCSF with 0.9% benzyl alcohol (pH 3.5). Data points are mean ± SD for triplicate samples. In some samples error bars represented are smaller than symbols.

Figure 2.

Loss of native rhGCSF due to aggregation followed as a function of time of incubation at 37°C by Method 1 (incubation of rhGCSF in a single eppendorf tube and analysis) and Method 2 (incubation of rhGCSF in separate eppendorf tubes for each time point and analysis). Upward triangles, rhGCSF in PBS (pH7.0)—Method 1; circles, rhGCSF in PBS—Method 2; downward triangles, rhGCSF at pH 3.5—Method 1; squares, rhGCSF in pH 3.5 (HCl)—Method 2. Data points are mean ± SD for triplicate samples. In some samples error bars represented are smaller than symbols.

Using Method 2, we compared the effects of 0.9% (w/v) benzyl alcohol on the aggregation of rhGCSF to that for the protein in buffer alone (Fig. 1). At pH 7.0 and 25°C, rhGCSF showed a progressive but relatively small loss of monomer over the 5-day incubation period, which was enhanced at 37°C. For samples incubated at pH 7 and 25°C, benzyl alcohol caused a relatively small increase in rhGCSF aggregation (Fig. 1A). In contrast, at pH 7 and 37°C, benzyl alcohol greatly accelerated aggregation. More than 90% protein aggregated within the first day, and there was complete loss of native protein by day 4 (Fig. 1B). The presence of 1.0 M sucrose partially counteracted benzyl alcohol-induced aggregation of the protein during incubation at both temperatures. In contrast to the observations in pH 7.0, at pH 3.5 there was essentially no aggregation of rhGCSF measured in the presence or absence of benzyl alcohol at either 25 or 37°C (Fig. 1A and B).

Effects of Benzyl Alcohol and Aggregation on Secondary Structure of rhGCSF

Second derivative IR spectroscopy was used to characterize the effects of benzyl alcohol on rhGCSF secondary structure prior to aggregation, and to characterize structural transitions in rhGCSF during aggregation in the presence or absence of benzyl alcohol. The predominantly α-helical structure of native rhGCSF in solution at both pH 3.5 and 7.0 (Figs. 3, 4, and 5), is reflected in the dominant band at 1656/cm16,19 At 25°C, prior to aggregation, 0.9% benzyl alcohol did not substantially alter the IR spectrum of the protein relative to that of the protein in buffer alone, at either pH 3.5 or 7 (data not shown). Thus, prior to aggregation, at 25°C benzyl alcohol did not perturb secondary structure of rhGCSF. Similar results were found at 37°C for samples at pH 3.5 (data not shown).

Figure 3.

Second derivative IR spectra of rhGCSF followed as function of time (up to 2.0 h) in different formulations at pH 7.0 and 25°C. (A) rhGCSF alone (B) rhGCSF with 0.9% benzyl alcohol. Solid line corresponds to rhGCSF liquid control at 25°C and pH 7.0. (C) Change in ratio of second derivative peak intensities 1620/1656 per cm as a function of time during kinetic scans at 25°C. Circles, rhGCSF alone (pH 7.0); downward triangles, rhGCSF and 0.9% benzyl alcohol (pH 7.0).

Figure 4.

Second derivative IR spectra of rhGCSF followed as function of time (up to 2.0 h) in different formulations at pH 7.0 and 37°C. (A) rhGCSF alone (B) rhGCSF with 1.0 M sucrose (C) rhGCSF with 0.9% benzyl alcohol (D) rhGCSF with 1.0 M sucrose and 0.9% benzyl alcohol. Solid line in all panels represents rhGCSF liquid control at 25°C and pH 7.0. (E) Change in ratio of second derivative peak intensities 1620/1656 per cm as a function of time during kinetic scans at 37°C and pH 7.0. Circles, rhGCSF alone; upward triangle, rhGCSF and 1.0 M sucrose; downward triangle, rhGCSF and 0.9% benzyl alcohol; squares, rhGCSF and 1.0 M sucrose and 0.9% benzyl alcohol.

Figure 5.

Second derivative IR spectra of rhGCSF followed as function of time (up to 2.0 h) at pH 3.5 in the presence of 0.9% benzyl alcohol. (A) At 25°C and (B) at 37°C. Solid line represents rhGCSF liquid control at 25°C and pH 3.5. (C) Change in the ratio of second derivative peak intensities 1620/1656 per cm as a function of time during kinetic scans at pH 3.5. Hexagons, rhGCSF and 0.9% benzyl alcohol at 25°C; cross-hair, rhGCSF and 0.9% benzyl alcohol at 37°C. Change in ratio for rhGCSF alone (diamonds) at 37°C and pH 3.5 is shown for comparison.

Next, to monitor secondary structural changes due to aggregation in real-time, we examined the IR spectra of rhGCSF as a function of time for 2 h. The IR spectra for the protein in buffer alone at pH 7.0 and 25°C showed no significant changes during this incubation, indicating that the protein was not aggregating (Fig. 3A and C). In the presence of 0.9% benzyl alcohol, the formation of a low intensity intermolecular β-sheet band at 1620/cm could be detected (Fig. 3B and C), indicating a small degree of aggregation.19 This effect was almost completely counteracted by 1 M sucrose at 25°C (data not shown). In the spectra for rhGCSF at 37°C and pH 7, the intensity of the aggregate band at 1620/cm increased rapidly with a concomitant decrease in the α-helical band intensity at 1656/cm (Fig. 4A and E), indicating extensive aggregation of the protein and loss of native secondary structure. Secondary structural transitions of rhGCSF in buffer at 37°C could not be detected in the presence of 1.0 M sucrose (Fig. 4B and E). Benzyl alcohol at 0.9% greatly enhanced the rate of the transition from the native to the aggregated state (Fig. 4C and E), with the appearance of a substantial aggregate band at 1620/cm in the first spectrum acquired (i.e., within about 80 s of mixing the protein with benzyl alcohol). Sucrose (1.0 M) substantially counteracted benzyl-alcohol induced acceleration of aggregation (Fig. 4D and E).

In the absence of benzyl alcohol at pH 3.5, at both 25 and 37°C, alterations in the secondary structure of rhGCSF were not detected during the 2-hour experiment (data not shown). Therefore even at the relatively high protein concentration (15 mg/mL) used for the IR spectroscopic studies, bands indicative of intermolecular β-sheet were not detected. Similarly, in the presence of 0.9% benzyl alcohol at 25°C there was not a detectable change in native secondary structure during the 2-hour experiment (Fig. 5A and C). However, at 37°C, the presence of benzyl alcohol induced some protein aggregation (Fig. 5B and C), as evidenced by the time-dependent decrease in the α-helical band intensity and concomitant increase in the intermolecular β-sheet band at 1620/cm. This result is in contrast to the absence of aggregation seen for this same sample during incubation in Eppendorf tubes at pH 3.5 and 37°C (Fig. 1B), and is likely due to a much higher protein concentration that was used for the IR spectroscopy studies (15 mg/mL, compared to 1.5 mg/mL used for incubation).

Under conditions where rhGCSF aggregated during the real-time IR spectroscopy experiments, there appeared to be a direct conversion from native α-helix (band at 1556/cm) to nonnative intermolecular β-sheet (band at 1620/cm), without a detectable intermediate secondary structural transition. Isosbestic points at approximately 1662 and 1640/cm (Figs. 3, 4, and 5) are consistent with a two-state structural transition from the conformation with native secondary structure to the aggregated state. We speculate that rhGCSF aggregation occurs via a partially unfolded intermediate8,24 and that this species has a secondary structure indistinguishable from that of the native protein and/or the intermediate may have an altered secondary structure and its levels are so low that it cannot be detected with IR spectroscopy (unpublished experiments).

Tertiary Structural Changes in rhGCSF Caused by Benzyl Alcohol Prior to Aggregation

The tertiary structure of rhGCSF was studied by fluorescence, second derivative UV absorbance and near UV CD spectroscopies prior to aggregation. rhGCSF contains two tryptophan and three tyrosine residues per monomer, which makes these methods sensitive to changes in its structure.13 Prior to studying effects of benzyl alcohol on the protein's spectra, we investigated the direct effect of the alcohol on fluorescence by using the model compounds N-acetyl-tryptophanamide and N-acetyl tyrosinamide in a 2:3 molar ratio. The results showed that 0.9% benzyl alcohol caused substantial quenching of fluorescence at pH 3.5 and 7.0 (Fig. 6A and B). Similar effects of benzyl alcohol on the solvent polarity were also observed through second derivative absorbance spectroscopy and resulted in change in peak positions of the model compounds in the presence of benzyl alcohol compared to samples in buffer alone (data not shown). Such effects of solvent polarity on peak shifts on the absorbance spectra due to the presence of organic solvent additives have been reported before.25 Because of the direct effects of benzyl alcohol on fluorescence and absorbance by aromatic amino acids, it was not possible to use these spectroscopic approaches to determine effects of the preservative on the tertiary structure of rhGCSF.

Figure 6.

Fluorescence spectra of control sample containing 2:3 ratio of N-acetyl-tryptophanamide and N-acetyl-tyrosinamide under different solution conditions upon excitation at 280 nm (A) pH 7.0: Solid-line, no benzyl alcohol, 25°C; medium dashes, with benzyl alcohol at 25°C; dash-dot-dot, no benzyl alcohol, 37°C; short dashes, with benzyl alcohol, 37°C. (B) pH 3.5: solid-line, no benzyl alcohol, 25°C; short dashes, with benzyl alcohol, 25°C; dash-dot-dot, no benzyl alcohol, 37°C; medium dashes, with benzyl alcohol, 37°C.

However, the effects of benzyl alcohol on tertiary structure of rhGCSF could be monitored by near UV CD spectroscopy. Near UV CD spectroscopy has been used extensively to monitor protein tertiary structure at different temperatures26,27 and with varying solvent conditions28,29 without any interference on the protein signal. Also, near-UV CD can be used reliably for understanding the direct effects of benzyl alcohol on the protein and in the absence of aggregation because of the low protein concentrations (1.0 mg/mL) required for the analysis and the relative speed of acquisition of spectra.

There was a decrease in the ellipticity of the near UV CD signal of rhGCSF in the presence of 0.9% benzyl alcohol compared to that for the protein in buffer alone at pH 7. This effect was more pronounced at 37 than at 25°C (Fig. 7A and B, Tab. 1). These results indicate that the asymmetric environment of the aromatic amino acid residues (Tyr, Trp) in the protein is altered in the presence of benzyl alcohol, and a greater degree of alteration occurs at 37 than at 25°C. No significant changes in ellipticity in the near-UV CD spectrum for the protein were seen in the presence of sucrose (1 M) at 25 and 37°C in pH 7.0 (Fig. 7A and B, Tab. 1). The effect of benzyl alcohol on the near-UV CD signal of the protein at pH 7.0 was partially reversed in the presence of sucrose at both temperatures (Fig. 7A and B, Tab. 1), suggesting partial counteraction of benzyl alcohol-induced perturbations of rhGCSF tertiary structure.

Figure 7.

Near-UV CD spectra of rhGCSF under different formulation and solution conditions at (A) 25°C and pH 7.0, (B) 37°C and pH 7.0. (C) 25°C and pH 3.5, (D) 37°C and pH 3.5. Solid line rhGCSF alone (pH 7.0); long-dashed line, rhGCSF with 1.0 M sucrose (pH 7.0); dotted line, rhGCSF with 0.9% benzyl alcohol (pH 7.0); dash-dotted line rhGCSF with 1 M sucrose and 0.9% benzyl alcohol (pH 7.0); small-dashed line rhGCSF alone (pH 3.5); dash-dot-dot line, rhGCSF with 0.9% benzyl alcohol at (pH 3.5).

Table 1. Near-Ultra Violet (UV) Circular Dichroism (CD) Signal Expressed as Mean Residue Ellipticity (MRE) for rhGCSF at 292 nm*
Samples25°C/pH 7.037°C/pH 7.025°C/pH 3.537°C/pH 3.5
  • *

    The calculated values represent mean ± SD for triplicate samples.

rhGCSF20.64 ± 0.119.68 ± 0.099.95 ± 0.108.59 ± 0.07
rhGCSF + Benzyl alcohol13.91 ± 0.175.01 ± 0.118.88 ± 0.107.32 ± 0.20
rhGCSF + Sucrose20.90 ± 0.2218.56 ± 0.19
rhGCSF + Sucrose + Benzyl alcohol17.04 ± 0.179.19 ± 0.12

At pH 3.5 in buffer alone there was an overall decrease in the CD signal intensity for the protein compared to the signal at pH 7.0. Such pH-dependent decrease in ellipticity has been observed previously for rhGCSF even though at pH values as low as 2 nativelike, compact structure is retained.30,31 The near-UV CD signal for rhGCSF in the presence of benzyl alcohol at pH 3.5 exhibited a smaller decrease (compared to effects at pH 7) in ellipticity at both temperatures (Fig. 7C and D, Tab. 1). Therefore, even though rhGCSF does not aggregate readily at pH 3.5 in the presence of 0.9% benzyl alcohol, the preservative perturbs protein tertiary structure at this pH.

Effects of Benzyl Alcohol on H–D Exchange in rhGCSF

The native state of a protein is an ensemble of conformational substates that include species ranging from those with a maximally compact native conformation to those that are partially unfolded.32,33 Across this range of conformations, the rate of exchange of protons for deuterium atoms increases as structural expansion increases. Therefore, relative H–D exchange rates for a protein provide information on the effects of solution conditions and other additives on the levels of partially unfolded species at any instant in time. From the viewpoint of behavior of an individual protein molecule, the results provide insight into the time-averaged conformation. The more compact that this conformation is, the greater the fraction of the protein's core that is “protected” from relatively rapid H–D exchange.34 Overall, a factor that increases the rate of H–D exchange does so by shifting the molecular population toward partially unfolded species, which is equivalent to an increase in the time-averaged partial unfolding of an individual protein molecule.

Real-time IR spectroscopy was used to investigate the effect of benzyl alcohol on H–D exchange of rhGCSF at 25 and 37°C, and pD's 6.6 and 3.1 (equivalent to pH's 7.0 and 3.5). With IR spectroscopy, as the protons in the peptide backbone of the protein are exchanged with deuterium, a loss in the NH stretching frequency is monitored as a change in the ratio of the amide II (centered around 1550/cm)/amide I (1600–1700) absorbances as a function of time (Fig. 8A).

Figure 8.

(A) Representative IR spectra of rhGCSF during deuteration in the presence of benzyl alcohol and sucrose at 25°C and pH 7.0, showing the amide I (1600–1700/cm) and amide II (1500–1600/cm) peaks. Ratio of Amide II/Amide I peak height followed a function of time during deuteration at pH 7.0. (B) 25°C and (C) 37°C. Circles rhGCSF alone; upward triangles, rhGCSF with 1.0 M sucrose; downward triangles, rhGCSF with 0.9% benzyl alcohol; squares, rhGCSF with 1 M sucrose and 0.9% benzyl alcohol. Each data point represents mean ± SE for triplicate samples.

The results for rhGCSF at pH 7.0 and at 25 and 37°C are shown in Figure 8B and C, respectively. The initial degree and extent of exchange after 120 min are greatly increased in the presence of 0.9% benzyl alcohol at 25°C, indicating that benzyl alcohol causes a shift in molecular population towards structurally expanded partially unfolded species. From the viewpoint of an individual protein molecule, it appears that the core of the native protein that is protected from rapid H–D exchange is much smaller in the presence of benzyl alcohol than in buffer alone; that is, the timed-average conformation of the protein is more structurally expanded. H–D exchange for rhGCSF in benzyl alcohol at 37°C and pH 7.0 could not be monitored due to the rapid aggregation of the protein under these conditions. This was confirmed by taking the second derivative of the absorbance signals for this sample, which showed intense aggregate bands at 1620/cm (intermolecular β-sheet) in the IR spectrum (data not shown). As seen in Figure 8B and C, the presence of sucrose (1 M) substantially lowers the exchange rate for the protein alone in buffer at 25 and 37°C in pH 7.0 indicating that in the presence of sucrose the overall molecular population is shifted towards more compact conformations.16,22,23 Further, sucrose partially counteracted the enhanced H–D exchange of rhGCSF in the presence of benzyl alcohol at the two temperatures studied.

At pD 3.1, benzyl alcohol increased the extent of H–D exchange for rhGCSF compared to that noted in buffer alone (Fig. 9) at both 25 and 37°C. The magnitude of this effect was greater at the higher temperature. These results document that benzyl alcohol shifts the molecular population of rhGCSF towards partially unfolded species at pH 3.5 and that this effect is enhanced at higher temperature, corroborating the observations from near-UV CD studies.

Figure 9.

Ratio of Amide II/Amide I peak height followed as a function of time of deuteration at pH 3.5. Circles, rhGCSF alone at 25°C; squares, rhGCSF alone at 37°C; upward triangle, rhGCSF with 0.9% benzyl alcohol at 25°C; downward triangle, rhGCSF with 0.9% benzyl alcohol at 37°C. Each data point represents mean ± SE for triplicate samples.

Effects of Benzyl Alcohol on Thermal Stability of rhGCSF

To study the effects of benzyl alcohol on the thermal stability of rhGCSF, the position of the tryptophan peak in the second derivative absorbance spectrum for rhGCSF and the optical density of the protein solution were measured as a function of temperature (Fig. 10A and B). Optical density (OD) in the 340–360 nm range (i.e., turbidity) is indicative of the presence of large protein aggregates.35 An increase in the wavelength minima for tryptophan with increase in temperature was seen under all solution conditions. However, due to extensive aggregation at pH 7.0, as evidenced by the large increases in OD at 350 nm, the data were truncated at 52°C for the protein in buffer alone and at 38°C in the presence of 0.9% benzyl alcohol.

Figure 10.

Thermal denaturation of rhGCSF in buffer alone and in the presence of 0.9% benzyl alcohol at pH's 3.5 and 7.0 as monitored using UV spectroscopy. (A) Change in tryptophan peak position as a function of temperature. (B) Optical density of rhGCSF at 350 nm as a function of temperature. Closed circle, rhGCSF alone (pH 7.0); open circle, (0.9% benzyl alcohol, pH 7.0); closed triangles, rhGCSF alone (pH 3.5); open triangles, (0.9% benzyl alcohol, pH 3.5); closed squares, 3:2 mixture of N-acetyl-tyrosinamide and N-acetyl-tryptophanamide in buffer alone (pH 7.0, PBS); open squares, 3:2 mixture of control sample in the presence of benzyl alcohol. Data points represent mean ± SD for triplicate samples.

In the presence of benzyl alcohol at pH 7.0, the temperature at which the turbidity initially increased was 38°C compared to 46°C in buffer alone. The decrease in turbidity observed at higher temperatures was due to precipitation of the protein out of the beam path in the instrument. Overall, these results document that benzyl alcohol reduces the stability of rhGCSF during heating because of induction of protein aggregation and are consistent with the benzyl alcohol-induced stimulation of aggregation measured during the isothermal incubations at 37°C.

At pH 3.5, at all temperatures studied, the tryptophan peak minima showed a red-shift in the presence of 0.9% benzyl alcohol compared to that for the protein in buffer alone, which could be due to effects of the alcohol on solvent polarity and/or changes in protein tertiary structure. However, during heating of both samples there also was a cooperative transition to higher wavelengths for the tryptophan minima. Approximately linear wavelength shifts were seen for the control samples consisting of N-acetyl tryptophanamide and N-acetyl tyrosinamide as a function of temperature, indicating that the transitions noted in samples with the protein are due to conformational changes in the protein and not due to direct temperature effects of tryptophan absorbance. The midpoint of the temperature-dependent transition for the protein was 69°C in buffer alone and 65°C in 0.9% benzyl alcohol (Fig. 10A). Thermal unfolding of the protein at this pH was only partially reversible (data not shown) probably due to the formation of small protein aggregates during heating that do not cause an increase in turbidity (Fig. 10B). Thus it is not certain whether the reduction in the apparent melting temperature of rhGCSF caused by benzyl alcohol at pH 3.5 is due to a decrease in the thermodynamic stability of the protein's native state and/or due to promotion of protein aggregation during heating.

DISCUSSION

Benzyl Alcohol-Induced rhGCSF Aggregation and It's Inhibition

Our results documented that 0.9% benzyl alcohol caused rapid aggregation and precipitation of rhGCSF at pH 7.0, associated with conversion of native α-helix structure into nonnative intermolecular β-sheet. In support of our hypothesis, benzyl alcohol also caused an increase in the level of highly aggregation-prone protein molecules with perturbed tertiary structure and native secondary structure, which are typical of the partially unfolded species that have been found to lead to aggregation of other proteins.23,36,37 Similar perturbations of protein tertiary structure by benzyl alcohol and a resulting acceleration of protein aggregation have been seen with recombinant human interleukin-1-receptor antagonist,8 a β-sheet protein. Furthermore, it has been found that other ligands, which promote protein aggregation, do so by increasing the level of partially unfolded protein molecules. For example, Congo red accelerates the aggregation of immuglobulin light chain variable domains by this mechanism.23 Thus, in general, solution conditions that cause partial unfolding of protein molecules can promote formation of nonnative aggregates.

A shift in equilibrium between molecular species towards partially unfolded protein molecules in the presence of benzyl alcohol and other ligands can be described by the Wyman Linkage function.38 According to this mechanism, preferential interaction of these aggregation-promoting ligands with the partially unfolded molecules would stabilize these species and result in an increase in their levels.8 Therefore, a rational strategy to counteract ligand-induced acceleration of protein aggregation is to employ solution additives that disfavor partially unfolded protein molecules. Sucrose is one such additive. Sucrose is preferentially excluded from the surface of protein molecules, which results in an increase in protein's chemical potential. The magnitude of preferential exclusion and chemical potential increment correlate directly with the surface area of the protein molecule exposed to solvent, and thus are greater for partially and fully unfolded molecules than for the most compact native state species. Therefore, sucrose shifts the equilibrium away from unfolded molecules, which should serve to counteract benzyl alcohol-induced acceleration of aggregation. In the current study we found that 1.0 M sucrose had both this effect and partially reversed the observed perturbations of rhGCSF tertiary structure by benzyl alcohol. Because sucrose's effects on protein stability are generally independent of protein properties,39 employing it as an agent to prevent preservative-induced protein aggregation could be widely applicable. Also, shifts in the molecular equilibrium towards the native state can be achieved with ligands that bind specifically to the native state of a given protein,40,41 which offers another strategy to inhibit aggregation of proteins caused by compounds that foster partial protein unfolding.

Effect of Temperature on Benzyl Alcohol-Induced rhGCSF Aggregation

Isothermal titration calorimetry was used in the earlier study on interleukin-1 receptor antagonist to document that benzyl alcohol interactions with the protein are hydrophobic.8 Furthermore, partially unfolded protein molecules usually have a higher degree of surface exposure of hydrophobic residues than do fully folded species,42 which accounts for preferential interactions of benzyl alcohol with these species. Because hydrophobic interactions are strengthened as temperature is increased, we hypothesized that benzyl alcohol-induced perturbations of rhGCSF structure and the resulting aggregation would be enhanced at 37°C compared to at 25°C. We were not able to measure the thermodynamics of benzyl alcohol binding to rhGCSF because the protein aggregated during isothermal titration calorimetry measurements. However, because of the strong temperature dependence of benzyl alcohol-induced structural perturbation and aggregation, our results are consistent with a predominant role for hydrophobic interactions in binding of benzyl alcohol to rhGCSF.

Other solution additives have been shown to interact with proteins hydrophobically and, hence, their effects on protein stability can be sensitive to temperature. For example, the thermal unfolding temperatures of proteins decreases in the presence of alcohols (e.g., methanol, ethanol, isopropanol) as the number of methyl groups in the alcohol are increased.43 But even alcohols such as ethanol can stabilize proteins against unfolding at low temperatures, such as those encountered during freeze-thawing. Similarly, nonionic surfactants are often employed to stabilize proteins against aggregation during freeze-thawing or during agitation at low temperatures (e.g., 4–8°C). This effect is often reversed at higher temperatures. For example, these compounds can decrease stability of proteins during thermal scanning, which is often used to screen effects of formulation components on protein stability.44 The important role of temperature in dictating the influence of a given compound on protein stability is particularly concerning for attempts in the biopharmaceutical industry to use high temperature as a means to accelerate protein degradation so that formulations can be screened rapidly. Additives such as nonionic surfactants, which might be effective and critical stabilizers for therapeutic protein products (which are typically stored at 4–8°C), could be rejected as formulation candidates because they reduce protein stability at elevated temperatures. Also, for in vitro high-throughput screening of ligands as potential therapeutic inhibitors of amyloid fibril formation in vivo, it is critical that experiments be conducted at 37°C and not at room temperature or in the refrigerator. Otherwise, compounds that could actually accelerate aggregation could be misindentified as protein stabilizers. Thus, it is crucial that the temperature dependency of the effects of additives on protein stability and the actual temperature of use be considered in the design of experiments to identify optimal protein formulations and inhibitors of amyloidosis.

Effect of pH on Benzyl Alcohol-Induced rhGCSF Aggregation

Unlike many other proteins, rhGCSF maintains a well-defined tertiary structure and highly helical secondary structure at pH values as low as 2, which makes it a good model protein for studying pH effects on preservative-induced protein aggregation.31 In addition to the distribution of species within the protein population, protein aggregation is governed by the energetics of protein–protein interactions.17 Previously, it has been shown that rhGCSF does not aggregate at pH 3.5, even though it aggregates rapidly at pH 7. The free energy of the protein unfolding is almost equal at both pH's, so conformational stability does not account for the difference in aggregation rates. However, at pH 3.5 rhGCSF is highly positively charged, resulting in strong protein–protein electrostatic repulsion. Conversely, at pH 7.0 intermolecular interactions are highly favorable.17 In our experiments with benzyl alcohol we found that such favorable energetics for intermolecular interactions at pH 7 allowed the structural perturbations in rhGCSF caused by benzyl alcohol to be manifested as acceleration of aggregation.

In contrast, the intermolecular repulsion between protein molecules at pH 3.5 greatly reduced benzyl alcohol-induced aggregation, even though the alcohol caused an increase in the level of partially unfolded, aggregation-prone species. Thus, an important strategy to inhibit ligand-induced protein aggregation is to maximize charge–charge repulsion between protein molecules (i.e., optimize colloidal stability) by the appropriate choice of pH. Of course, for this strategy to be effective it is important to determine a pH that does not by itself lead to sufficient conformational perturbations to overcome the beneficial effects of intermolecular repulsion. Overall, it is usually not possible to predict for a given protein the effects of pH on the balance between conformational and colloidal stabilities. Therefore, it is necessary to determine empirically the optimal pH for minimizing protein aggregation.

CONCLUSIONS

The results from the present study further document that benzyl alcohol accelerates protein aggregation by favoring the formation of partially unfolded, aggregation-prone protein species in the protein molecular population. The similarity of this mechanism for a predominantly α-helical (rhGCSF) and a β-sheet protein8 suggests that it is commonly applicable to most proteins. As evident from this study, effective control of preservative induced aggregation of protein molecules can be achieved by modulating solution conditions and rational choice of excipients. Inclusion of preferentially excluded additives such as sucrose can counteract preservative-induced protein aggregation by favoring the most compact conformation in the protein molecular population. Nonnative protein aggregation also can be inhibited by maximizing colloidal stability through the testing for a pH that optimizes intermolecular charge–charge repulsion.17 These strategies should be broadly applicable to the inhibition of protein aggregation in general, as well as to the specific problem of ligand-induced protein aggregation.

Acknowledgements

This research was supported by grants from NSF (BES-0138595) and Amgen, Inc. We gratefully acknowledge Pharmaceutics Department, Amgen, Inc., for providing us with technical support for the project.

Ancillary