Freezing-Induced Phase Separation and Spatial Microheterogeneity

Jul 2, 2009 - Note that the laser that was used in Raman spectroscopy is of low power. (∼2mW) and ... to what was observed in the HS samples. Howeve...
0 downloads 0 Views 2MB Size
J. Phys. Chem. B 2009, 113, 10081–10087

10081

ARTICLES Freezing-Induced Phase Separation and Spatial Microheterogeneity in Protein Solutions Jinping Dong,†,§ Allison Hubel,‡ John C. Bischof,‡ and Alptekin Aksan*,‡ Characterization Facility, Institute of Technology, UniVersity of Minnesota, Minneapolis, Minnesota 55455, and Mechanical Engineering Department, 111 Church Street Southeast, UniVersity of Minnesota, Minneapolis, Minnesota 55455 ReceiVed: NoVember 3, 2008; ReVised Manuscript ReceiVed: May 11, 2009

Amid decades of research, the basic mechanisms of lyo-/cryostabilization of proteins and more complex organisms have not yet been fully established. One major bottleneck is the inability to probe into and control the molecular level interactions. The molecular interactions are responsible for the significant differences in the outcome of the preservation processes.1 In this communication, we have utilized confocal Raman microspectroscopy to quantify the freezing-induced microheterogeneity and phase separation (solid and liquid) in a frozen solution composed of a model protein (lysozyme) and a lyo-/cryoprotectant (trehalose), which experienced different degrees of supercooling. Detailed quantitative spectral analysis was performed across the ice, the freeze-concentrated liquid (FCL) phases, and the interface region between them. It was established that the characteristics of the microstructures observed after freezing depended not only on the concentration of trehalose in the solution but also on the degree of supercooling. It was shown that, when samples were frozen after high supercooling, small amounts of lysozyme and trehalose were occluded in the ice phase. Lysozyme preserved its native-like secondary structure in the FCL region but was denatured in the ice phase. Also, it was observed that induction of freezing after a high degree of supercooling of high trehalose concentrations resulted in aggregation of the sugar and the protein. Introduction The worldwide market for the 75 FDA approved therapeutic proteins is expected to reach $70 billion by the end of 2008. Currently, approximately 500 therapeutic proteins are in the development or approval stages to be used in the treatment of various conditions ranging from cancer (e.g., monoclonal antibodies and interferons),2 heart attack and stroke (e.g., growth factors and antiplatelet factors),3 anemia (erythropoietin to improve red blood cell production),4 and hemophilia (blood clotting factor VII).5 Proteins also find widespread use as biocatalyzers, and in bioreactors and biosensors.6-9 The very high costs associated with the production, isolation, and purification of the therapeutic proteins necessitate that the proteins should be successfully stabilized at high concentrations, and stored with minimum loss of activity.10-12 The majority of the therapeutic proteins are stabilized and preserved by drying, freezing, or freeze-drying.13-17 Stabilization of a protein by freezing and freeze-drying involves addition of buffering and stabilization/preservation agents into the solution,16,18,19 and removal/separation of the liquid water from the solution in the form of ice. The presence of chemical agents in the solution and the freezing/freeze-drying processes change the chemical potential of water, which directly alters protein structure.20 The structure of the protein is directly related to its activity after * Correspondingauthor.E-mail:[email protected]:612.626.6618. Fax: 612.624.5230. † Characterization Facility. ‡ Mechanical Engineering Department. § Present address: Characterization Facility, Institute of Technology, 12 Shepherd Laboratory, 100 Union St. S.E., Minneapolis, MN 55455.

reconstitution. Aggregation, gelation, and cold-denaturation are examples of processing and storage-induced physical changes that result in the loss of protein stability, and therefore loss of its activity.11,21 Freezing of water induces solute rejection, creating regions of high solute concentration. Freezing-induced partitioning of the solution into different thermodynamic phases (an ice phase and a freeze-concentrated liquid (FCL) phase) induces segregation of the protein, exposing it to different microenvironmental conditions within the same medium. The local microenvironment that the protein is exposed to continues to evolve as the medium is cooled further and eutectic or crystalline phases form or the FCL region vitrifies. A similar microsegregation phenomenon is also observed during room-temperature desiccation of protein solutions, even though the physical mechanisms are different.22 Phase separation and microsegregation in protein solutions during desiccation, freezing,23 and lyophilization24 has been previously studied in bulk solutions, mainly by thermal, kinetic, and calorimetric analyses,12,25-30 X-ray diffractometry,31 and infrared (IR) spectroscopy30,32,33 of the bulk product. The ultrastructure of the processed products has been visualized using scanning electron microscopy (SEM).34,35 However, it is yet not known how the protein is distributed among the different regions in the frozen medium, and how its structure is affected by the freezing-induced partitioning and temperature dependent evolution of these different regions.12,13 Simply, the question we set to answer is: “In a frozen medium, is the protein even “seeing” the stabilization agent to benefit from its protection?” Previous studies used SEM or

10.1021/jp809710d CCC: $40.75  2009 American Chemical Society Published on Web 07/02/2009

10082

J. Phys. Chem. B, Vol. 113, No. 30, 2009

Dong et al.

Figure 1. CRM images of frozen 1×PBS solution (HS: high degree of supercooling). Scale bar ) 5 µm.

electron spectroscopy for chemical analysis (ESCA) to answer this question in dried and freeze-dried protein formulations and demonstrated that protein and the stabilizing agents frequently do not colocalize.17,22 In this study, we employed confocal Raman microscopy (CRM) with frozen protein formulations to obtain 3-D high spatial resolution (∼200-300 nm) chemical maps of the system to quantify the extent of microsegregation of proteins into the coexistent thermodynamic phases formed during freezing. We measured the spatial distribution of the excipients in the frozen medium, identified the conditions that yield to aggregation, and determined the changes in the secondary structure of the model protein as a function of the freezing protocol applied and the composition of the medium. Results and Discussion Isotonic Saline Solution. Buffers such as phosphate buffered saline solution (PBS), divalent cations (Zn2+), as well as basic amino and dicarboxylic acids have commonly been used in the cryo-/lyopreservation formulations developed for proteins, vaccines, as well as micro-organisms and mammalian cells.16,19,36,37 However, it is also known that sodium and potassium-based buffers, upon freezing, may destabilize proteins38,39 due to crystallization of the buffer and the significant change of pH. Therefore, one should be extremely cautious when using a buffer with a high concentration of sodium-based salts, although, in practice, PBS is used very commonly in lyophilized biopharmaceutical products.16 Solidification patterns of the PBS solution are complex (Figure 1) and continue to evolve with time as ice crystals coalesce and grow (white arrows in Figure 1). Two characteristic peaks located in the ν-OH band (2600-3600 cm-1) of the Raman spectra were used to identify the thermodynamic phase of the water (solid vs liquid) within the frozen sample at very high spatial resolution (lateral plane, ∼200-300 nm; vertical plane, ∼500-600 nm). The peak characteristic of liquid water was centered at 3430 cm-1, and the sharp peak located at 3145 cm-1 corresponded to ice.40 The spatial change in the ratio of the integrated areas under these two peaks was used to generate the 2-D CRM images that were used in the analysis. Note that the laser that was used in Raman spectroscopy is of low power (∼2mW) and the integration time (laser exposure at each pixel scanned) was less than 0.1 s for image acquisition and less than 1 s for spectral analysis. Therefore, heating of the sample by the laser was not expected. However, we also conducted experiments where a single spot inside the ice domain of the sample was exposed to prolonged irradiation (>10 min). No spectral changes indicative of temperature or phase change effects were observed. Note that the spectral characteristics of the Raman peaks of water change with temperature and also

with phase (ice spectrum is different than that of liquid water at the same temperature).41 However, we did not observe any change during prolonged exposure studies. In the spontaneously frozen 1×PBS solution (i.e., high supercooling (HS), with Tfreezing ∼ -20 °C), large ice crystals (daverage ∼ 10-20 µm) with a relatively uniform size distribution were observed (Figure 1). The smooth, round boundaries of the ice crystals indicated spherulitic morphology characteristic of the hexagonal ice phase formed under equiaxed freezing conditions.42 The crystals were surrounded by narrow channels and lacunae that were devoid of ice (i.e., the FCL regions). FCL regions were formed as a result of the depression of the freezing temperature of the solution, which now had a higher solute content than the original solution due to the presence of the solutes rejected by the nucleating and growing ice phase. When the frozen 1×PBS solution was held at a constant temperature (T ) -26 °C), the ice and the FCL regions rearranged slowly with time: ice crystals coalesced (white arrows in Figure 1), while the distinct FCL regions decreased in number but increased in size. The recrystallization process continued for approximately 60 min at T ) -26 °C. The 1×PBS solutions manually nucleated at T ) -3 °C (freezing with low supercooling (LS)) and cooled down to T ) -26 °C (dT/dt ) 10 °C/min) showed phase separation similar to what was observed in the HS samples. However, the ice crystal size distribution was much broader (daverage ∼ 2-25 µm). Luyet and Rapatz42 report similar observations that in frozen aqueous solutions the resulting crystalline microstructures are influenced by the composition of the solution, thickness of the solution layer, and the freezing temperature. Low Trehalose Concentration Lysozyme Solutions (100 mg/mL TRE:20 mg/mL LYS). For the majority of proteins, successful stabilization (by freezing, freeze-drying, vitrification, or desiccation) requires the use of specific stabilizing agents.10-12 These agents preserve the protein structure, eliminate/minimize aggregation, and ensure protein activity after reconstitution.13-15,43 Sugars are commonly used to improve protein stability,44,45 and there has been considerable interest specifically in the nonreducing disaccharide trehalose (TRE) as a stabilizing agent.46-48 The thermodynamic mechanism of stabilization offered by sugars such as TRE has been established.49 Recent studies have demonstrated that the addition of TRE to a conventional cryopreservation solution can change the interfacial free energy of the ice/cell/liquid system and influence partitioning of cells with respect to the solid phase.50 In order to quantify the effect of TRE on the solution freezing behavior and phase separation, 1×PBS solutions containing TRE and a model protein, lysozyme (LYS), were either allowed to spontaneously freeze during cooling (HS)

Freezing-Induced Phase Separation

J. Phys. Chem. B, Vol. 113, No. 30, 2009 10083

Figure 2. CRM images and line scan analyses of frozen TRE + LYS in 1×PBS solution. Column A, 20 mg/mL LYS + 100 mg/mL TRE solution (HS); column B, 20 mg/mL LYS + 100 mg/mL TRE solution (LS); column C, 20 mg/mL LYS + 300 mg/mL TRE solution (HS); column D, 20 mg/mL LYS + 300 mg/mL TRE solution (LS). Row 1: CRM images of the samples. Row 2: Spatial variation of ice concentration (dashed line), TRE hydration (solid blue line), and LYS/TRE concentration ratio (solid red line). Row 3: Spatial variation of LYS R-helix (solid red line) and β-sheet (solid blue line) content, and LYS hydration (solid green line). Note: The white line in the first image shows the location of the line scan in the first sample (HS, high degree of supercooling; LS, low degree of supercooling).

down to T ) -26 °C (dT/dt ) 10 °C/min) or initially seeded with ice at T ) -3 °C (LS) and then cooled down to T ) -26 °C (dT/dt ) 10 °C/min). Spontaneously frozen (HS) TRE-LYS solutions formed irregular ice crystals with a broad size distribution (daverage ∼ 2-20 µm; Figure 2, A1). The total concentration of the organic matter (TRE + LYS) in different regions within the frozen samples was determined using the integrated intensities of a group of peaks located at 2900-3000 cm-1 (Figure 3), which correspond to the ν-CH bands originating both from the protein and the sugar.51 In the long tortuous channels separating the ice domains (FCL region), the liquid water peak (3430 cm-1) showed very low contrast relative to the ν-CH peak (Figure 2, A1), indicating the presence of a very low concentration of liquid water in this phase as compared to the unfrozen solution. Note that, if we assume that the phase diagram of 8% w/w NaCl solution is similar to that of 1×PBS, in the presence of >100 mg/mL TRE at T ) -26 °C, under equilibrium conditions, there would only be ice and FCL phases present in the system, which would be devoid of any crystallized salts. In addition to the tortuous channels separating the ice domains, there were round lacunae located within the ice domains where high concentrations of TRE and LYS accumulated (white arrows in Figure 2, A1). It was determined by scanning several layers in the vertical direction that the round inclusions were very small (as compared to the tortuous channels) with a vertical dimension of 1-2 µm (which was

Figure 3. Typical Raman spectra collected in the ice, FCL, and ice/ FCL interface regions in a frozen 20 mg/mL LYS + 100 mg/mL TRE solution (HS).

similar to their lateral dimensions), suggesting that they had spherical geometry. In the HS TRE-LYS samples, time dependent coalescence of ice crystals, which was observed in the frozen 1×PBS solution, was absent and no significant change was observed in the microstructure over a 60 min period of time. However, in some prolonged experiments (>4 h), it was found that the tortuous channels slowly broke into smaller segments and eventually transitioned to small lacunae. A high resolution Raman scan across the tortuous channel between the two ice regions (along the white line shown in

10084

J. Phys. Chem. B, Vol. 113, No. 30, 2009

Dong et al.

TABLE 1: Average Concentrations of TRE and LYS in Different Phases concentration (mg/mL) solution 100 mg/mL TRE + 20 mg/mL LYS

freezing condition HS LS

300 mg/mL TRE + 20 mg/mL LYS

HS LS

solute

FCL region

ice region

Ka

LYS TRE LYS TRE LYS TRE LYS TRE

133.1 705.6 143.7 778.6 40.1 729.3 51.3 790.6

3.1 9.5 4.7 16.1 9.2 68.9 3.9 47.3

0.023 0.0135 0.033 0.021 0.23 0.09 0.078 0.06

a The partition coefficient; K ) concentration in ice/concentration in FCL.

Figure 2, A1) was performed in the HS samples to analyze the compositions of the ice and the FCL regions as well as the interface between them (Figure 3). The spatial distribution of LYS with respect to TRE in different regions was assessed by measuring the ratio of the integrated area of the tryptophan peak located at 1552 cm-1 to that of the C-C-C ring deformation peak of TRE located at 536 cm-1.52 It was determined that the ice regions (even outside the lacunae) were not composed of pure water but contained low concentrations of LYS and TRE. This was evident from the characteristic peaks of LYS and TRE observed in the spectra obtained in the ice region (Figure 3). Additionally, it was observed that the LYS/TRE ratio was 2-4 times higher in the ice phase (Figure 2, A2), which indicated that, during ice nucleation and growth, the ice phase preferentially rejected more TRE than LYS. Entrapment of LYS and TRE in the ice phase is very unusual. Under equilibrium conditions, the partition coefficient (defined as the concentration of a given solute in the solid phase divided by the concentration of a given solute in the liquid phase) of most solutes in the ice phase is extremely small (∼10-6, see Korber et al.53 for a review). On the other hand, the presence of detectable levels of LYS and TRE in ice is consistent with rapid ice interface propagation, resulting from the high supercooling.54 Rapid propagation of the solid/liquid interface has been associated with an increase in the partition coefficient and, correspondingly, an increase in the concentration of solutes (in this case, LYS and TRE) in the solid phase.55 A previous study also showed that the inclusion of solutes in the ice was dependent on the ice growth rate and the direction of the thermal gradient. Even though it was not possible by CRM analysis to precisely determine how TRE and LYS molecules were distributed in the ice phase, it is highly likely that both solutes were occluded in the ice phase in very small lacunae. Generally AgCl or similar molecules are capable of being incorporated into the ice lattice,56 and the complex structures and large sizes of LYS and TRE make integration into the ice lattice unlikely. To quantify the absolute concentrations of TRE and LYS in each region (FCL and ice phases), the changes in the peak intensities at 1552 cm-1 (for LYS) and 536 cm-1 (for TRE) with concentration were calibrated using spectra obtained from binary and ternary solutions containing 20-400 mg/mL TRE and/or 20-100 mg/mL LYS. The LYS/TRE ratio in each phase was calculated using area-averaged spectra collected from the FCL and ice regions. The volume of each phase was estimated from the areas in the 30 × 30 µm2 CRM images (first row in Figure 2). Table 1 lists the average concentrations of TRE and LYS in the ice and FCL regions for all conditions tested. In the 100 mg/mL TRE HS samples, the concentration of TRE in the

FCL region was 705.6 mg/mL (7 times its concentration in the solution), while LYS was 133.1 mg/mL (approximately 6.7 times more concentrated in FCL). The measured concentration of TRE in FCL was in the lower end of the range (705-833 mg/mL) reported in the literature for frozen TRE solutions.57 This was attributed to the presence of LYS in our experimental system. In the ice phase, the TRE concentration was about 1/10 of its concentration in the solution, while LYS was about 1/7 less concentrated (Table 1). Depending on their immediate microenvironment (i.e., entrapped in the ice region or in the FCL region), TRE and LYS presented significantly different spectral features than those in solution. The major differences were observed at the two doublet peaks located at 1440-1460 and 1060-1080 cm-1, which corresponded to the δ-CH2 and the combination ν-C-O, ν-C-C, and δ-COH vibrations, respectively.58 The doublet at 1450 cm-1 has contributions from both TRE and LYS; therefore, the relative intensities of the Gaussian peaks in the doublet located at 1070 cm-1 were used to quantify the hydration level of TRE. There are different definitions of the “hydration level” used in the literature. The definition we adopted is the one offered by Kacˇura´kova´ and Mathlouthi,58 where the hydration level is evaluated using the ratio of the 1080 cm-1 peak (ν-C-C and δ-COH) to the 1060 cm-1 peak (ν-C-O). In a separate study, we have also confirmed the hydration sensitivity of these doublets using FTIR spectroscopic analysis of concentrated sugar solutions at low temperatures.59 Figure 2, A2 (blue line, solid circles), shows the variation in TRE hydration (degree of hydrogen bonding) across the ice and FCL phases (ice concentration is shown as the dashed line). Overall, TRE was approximately 3 times more hydrated in the FCL region than in the ice phase. The high hydration level of TRE in the FCL region combined with the observation of a lower LYS/TRE ratio in the same region point to three distinct possibilities: (1) preferential binding of TRE to the remaining unfrozen water molecules (which is in accord with the preferential exclusion hypothesis,60,61 (2) increase in the TRE-TRE intermolecular interactions in the FCL region, and/or (3) preferential binding between the TRE and LYS (which is in accord with the water replacement hypothesis62). To determine the responsible phenomena, we quantified the changes in the secondary structure of LYS within different regions. The peak at ca. 1260 cm-1 in the amide III region originates from R-helical structures, while the peak at ca. 1238 cm-1 originates from β-sheet structures.52,63 Figure 2, A3, shows the variation in the R-helix (the red line, the open squares) and the β-sheet (the blue line, the full triangles) content of LYS across the different regions (note that the ice concentration profile is the same as that shown by the dotted line in Figure 2, A2). With reduction in the hydration level of LYS (as seen in the ice region), the R-helix content decreases, while the β-sheet and the random coil contents increase.63,64 The R-helix and β-sheet contents of the LYS changed considerably between different regions, with the R-helix content reaching its maximum (showing native-like secondary structure) in the FCL region. The corresponding drop in the β-sheet content in the FCL region supported this observation. Increase in the β-sheet content of LYS during freezing and desiccation was confirmed in parallel experiments performed with FTIR spectroscopy (data not shown) and therefore was attributed to the loss of native-like structure. The native-like structure of LYS in the FCL region could be attributed to the absence of ice, and a lower LYS/ TRE ratio, and possibly to a specific interaction of TRE with

Freezing-Induced Phase Separation

J. Phys. Chem. B, Vol. 113, No. 30, 2009 10085

Figure 4. (A, B) CRM images of frozen 20 mg/mL LYS + 300 mg/mL TRE solution (HS). (C) High resolution CRM image (the white square in image B). (D) Line scan (across the black line in image C) of the same solution showing aggregation of TRE and LYS in the FCL region.

LYS (by either exclusion or binding), preserving its structure. Previous reports in the literature show that LYS secondary structure is relatively stable during temperature change but is sensitive to the change in the hydration state of its side chains.65,66 To further quantify the difference in the hydration levels of LYS in different regions, hydration of LYS was analyzed by plotting the intensity ratio of the peaks at 1032 cm-1 (the phenyl ring of the phenylalanyl residue) and 1013 cm-1 (the tryptophanyl residue).52 The higher the intensity ratio of the peak located at 1032 cm-1 to that at 1013 cm-1, the higher the hydration level of LYS.67 The analysis revealed that LYS was 2-3 times more hydrated in the FCL phase than in the ice phase (green line in Figure 2, A3, full circles). In the LS samples cooled down to T ) -26 °C after manual ice seeding at -3 °C (Figure 2, B1), the ice crystals were very uniform, and exhibited a plate-like geometry with a high degree of anisotropy (with dimensions of 5-10 µm by 10-20 µm in the x-y plane and much smaller dimensions in the vertical plane). These structures were stable and did not exhibit significant coarsening/coalescence over a 60 min period when the samples were kept at T ) -26 °C. As expected, TRE and LYS were concentrated in the FCL phase between the ice crystals. In the ice phase, the absolute concentrations of both LYS and TRE were about 1.5 times higher than those measured in the HS samples (Table 1). These results showed that more LYS and TRE were entrapped in the ice phase in LS samples than HS. This is counterintuitive, since increasing the temperature at which the solution nucleates reduces the interface velocity of the ice crystal during the initial nucleation/crystal growth phase, reducing the effective partition coefficient of the solutes. However, in this case, the partition coefficient increased with lower supercooling. In the LS samples, LYS structural variation across different regions (red and blue curves in Figure 2, B3) followed a similar trend as the HS samples with a nativelike structure in the FCL region. The only significant difference between the two freezing protocols was that the overall hydration level of LYS was lower in the LS samples (green curve in Figure 2, B3). High Trehalose Concentration Lysozyme Solutions (300 mg/mL TRE:20 mg/mL LYS). In order to quantify the effects of increasing the sugar concentration in the solution, freezing experiments were carried out with sugar/protein solutions that contained a higher concentration of TRE (300 mg/mL). In the HS samples, relatively uniform and plate-like ice crystals were observed. However, these crystals were much smaller in size (∼2 × 5 µm, Figure 2, C1) than those obtained in the low TRE concentration solutions (∼5 × 15 µm, Figure 2, B1). Contrary to the relatively homogeneous distribution of TRE and LYS in the FCL region that was observed in the low TRE solutions, at

high TRE concentrations, small aggregates of pure organic matter were observed in the FCL region (white arrows in Figure 4B). Analysis at a higher spatial resolution (Figure 4C) showed that these aggregates containing TRE and LYS preferentially concentrated near the ice/FCL interface (white arrows in Figure 4C). Note that Figure 4D shows the total solute (TRE + LYS) concentration profile along the line shown in Figure 4C. Aggregation of LYS indicates a higher degree of denaturation, which is often irreversible due to the presence of high levels of non-native, intermolecular β-sheet structures.68 Accumulation of solutes at the ice/FCL interface has previously been observed by transmission electron microscopy of bacteria in a freezedried substrate.69 Similar to those observed in the low TRE solutions, the LYS/ TRE ratio (red curve in Figure 2, C2) was higher in the ice phase (Table 1). The hydration levels of TRE in both the ice and FCL phases were lower than those in the low TRE concentration cases (blue curve in Figure 2, C2). This finding implied that TRE hydration depended mainly on its specific interactions with LYS. The protein structure in the ice phase had significantly higher β-sheet content (blue curve in Figure 2, C3) even though the LYS/TRE ratio was lower (more TRE molecules should be available per LYS molecule). Similarly, the hydration level of LYS was lower in the ice and higher in FCL regions (green curve in Figure 2, C3). As compared to the high TRE HS samples, in the high TRE LS samples, larger (4 × 15 µm) plate-like ice crystals with a more uniform distribution were observed (Figure 2, D1). However, the distribution of TRE and LYS in the FCL region was more uniform without any significant aggregation at the ice/FCL interface (red curve in Figure 3, D2). The LYS/TRE distribution was significantly different with a higher concentration found in the center of the ice phase, which gradually decreased toward the interface. The more gradual change in the LYS/TRE concentration across all regions measured in the LS samples could be attributed to the functional difference in the temperature dependence of the partition coefficients of the two solutes. Conclusion Solidification Microstructures. The solidification microstructures observed in the presence of TRE and LYS exhibited a high degree of anisotropy in contrast to the spherulitic morphology observed in the frozen 1×PBS solutions. Low TRE concentration solutions that were allowed to freeze spontaneously (HS) had the most irregular ice shapes (non-plate-like structures) and a wide size distribution. As the TRE concentration in the solution increased, the ice crystal size became smaller; the ice crystals became plate-like with a more uniform shape

10086

J. Phys. Chem. B, Vol. 113, No. 30, 2009

and size distribution. Previous studies have shown similar phenomena of lyophilization-induced phase separation using SEM.35 Purity and Homogeneity of the Phases. The ice phase was not pure, but low levels of TRE and LYS were present in the ice crystals either in the form of large lacunae or smaller inclusion bodies. The LYS/TRE distribution was relatively uniform in the FCL regions (except in the 300 mg/mL TRE HS solution where organic aggregates formed). However, the distribution of TRE and LYS hydration was not uniform, reaching their highest levels at the center of the tortuous FCL channels. This might be attributed to the relatively heterogeneous distribution of water even in the FCL region (see the ice profiles in Figure 2, A2-D2). Note also that the LYS/TRE distribution was significantly lower in the high TRE samples across all regions. With freezing, the ice phase preferentially rejected solutes; however, the concentration of TRE entrapped in the ice phase was higher in all of the solutions tested. Since the size of TRE is much smaller than LYS, this could only be explained by preferential ice-LYS interactions. It is known that LYS has a very high surface affinity toward air/liquid interfaces.70 Its surface adsorption kinetics are also affected by the specific sugar type and its concentration in the solution.71,72 A similar affinity for the ice interface might have caused the increased LYS concentration within the ice region. Hydration of Molecules and Structure. The hydration level of TRE in the FCL phase decreased with increasing TRE concentration in the solution (Figure 2, A2-D2). Presumably, at high concentrations, TRE preferentially interacted with LYS (which was obvious in the spontaneously frozen samples forming the aggregates) to satisfy hydrogen bonding requirements. The higher native-like structural content of LYS in the FCL region also shows that TRE interacts more with LYS, preserving its structure. A direct correlation among TRE and LYS hydration levels and the R-helix content of LYS strengthens this conclusion. In this research, we have utilized CRM with frozen solutions to determine the effects of freezing kinetics on the segregation behavior of proteins among different coexisting thermodynamic phases. Enhancing our understanding of phase separation and the effects of the interactions between the protein and the ice interface on protein structure and aggregation will permit us to improve the stabilization of proteins. The findings presented in this research open new avenues for exploration not only to understand the mechanisms of biostabilization but also for protein isolation and purification processes and development of highly concentrated, stable therapeutic protein suspensions (which currently presents a road block for the medical advances based on protein therapies73). Materials and Methods Hen egg white lysozyme was purchased from Sigma (SigmaAldrich Corp., St. Louis, MO). High purity trehalose dihydrate was purchased from Pfanstiehl (Ferro Pfanstiehl Laboratories Inc., Waukegan, IL). Other chemicals were purchased from Sigma and used as received. Confocal Raman microspectroscopy (CRM) was conducted with a WITec Alpha 300R confocal Raman microscope (WITec Instruments Corp., Germany). The Raman microscope was equipped with a UHTS200 spectrometer and a DV401 CCD detector. A 100× Nikon air objective (NA ) 0.90) was used for all measurements. A 514.5 nm AR-ion laser at a maximum power of 50 mW (operated at 2 mW) was used for excitation.

Dong et al. High throughput optics of the microscope enabled 3-D confocal chemical mapping at a spatial resolution of 150-300 nm laterally and 500 nm vertically. A cold water cooled Peltier stage (Agilent Technologies Inc., Tempe, AZ) was used for controlled rate cooling of the samples. A 1 µL portion of the experimental solution was placed on an aluminum surface, and a small piece of thin polystyrene (PS) film (∼2 µm) was placed on top of the liquid droplet to prevent evaporation or sublimation during freezing and data acquisition. The sample and the Peltier stage were mounted on a piezoelectric scanning table placed under the Raman microscope. The experimental solutions were either cooled down to T ) -26 °C at a rate of dT/dt ) 10 °C/min with manual ice seeding at T ) -3 °C (low supercooling, LS) by touching the sample with a liquid N2 chilled needle or allowed to spontaneously nucleate in the range -20 °C < T < -26 °C (high supercooling, HS). 3-D CRM images were collected at T ) -26 °C at different depths (∼2-6 µm) below the PS film by raster scanning. An array of spectra (e.g., 80 × 80 corresponding to a sample area of 30 × 30 µm2) was collected using identical integration times at each pixel. CRM images were generated by integrating one or more of the characteristic spectral peaks. Calibration experiments were conducted with solutions of varying TRE and LYS content to validate that the tryptophan and the glycosidic link band peaks used in the analysis are not dependent on the level of hydration but concentration. We have also carefully calibrated our measurements and assessed the error in our concentration measurements as follows: Raman chemical maps were collected in 25 µm × 25 µm areas (80 pixels by 80 pixels) in pure TRE or LYS solutions of different concentration. The concentration of the solute (TRE or LYS) was calculated at each pixel using the characteristic Raman peak. The standard deviation of the measurement was calculated on the basis of the values obtained in each pixel in the homogeneous solution. It was found that the measurement error (i.e., % standard deviation from the average) was inversely correlated to concentration. This was expected, since at lower concentrations the Raman peak size was smaller. For TRE solutions, the error was below 1% when the TRE concentration was higher than 700 mg/mL, and was about 15% when the concentration was below 10 mg/mL. Acknowledgment. This research was supported by a grant from the Institute for Engineering in Medicine at the University of Minnesota. Supporting Information Available: Raman spectra obtained from pure solutions (100 mg/mL TRE in 1×PBS and 20 mg/ mL LYS in 1×PBS) at 20 and -26 °C. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Roy, I.; Gupta, M. N. Biotechnol. Appl. Biochem. 2004, 39, 165– 177. (2) Schrama, D.; Reisfeld, R. A.; Becker, J. C. Nat. ReV. Drug DiscoVery 2006, 5, 147–159. (3) Ken’Ichiro, K.; Shigeo, O.; Yasuo, K. Neurol. Ther. 2005, 22, 219– 224. (4) Johnston, J.; Tazelaar, J.; Rivera, V. M.; Clackson, T.; Gao, G.-P.; Wilson, J. M. Mol. Ther. 2003, 7, 493–497. (5) Smales, C. M.; James, D. C. Therapeutic Proteins: Methods and Protocols; Humana Press: Totowa, NJ, 2005. (6) Bashir, R. AdV. Drug DeliVery ReV. 2004, 56, 1565–1586. (7) Bjerketorp, J.; Hakansson, S.; Belkin, S.; Jansson, J. K. Curr. Opin. Biotechnol. 2006, 17, 43–49. (8) Bloom, F. R. Biosens. Bioelectron. 2001, 16, 603–608. (9) Lyngberg, O. K.; Stemke, D.; Schottel, J. L.; Flickinger, M. C. J. Ind. Microbiol. Biotechnol. 1999, 23, 668–676.

Freezing-Induced Phase Separation (10) Chang, L.; Shepherd, D.; Sun, J.; Ouellett, D.; Grant, K. L.; Tang, X.; Pikal, M. J. J. Pharm. Sci. 2005, 94, 1427–1444. (11) Ragoonanan, V.; Aksan, A. Transfus. Med. Hemotherapy 2007, 34, 246–252. (12) Randolph, T. W. J. Pharm. Sci. 1997, 86, 1198–1203. (13) Pikal, M. J. In Stability, Formulation and DeliVery of Peptides and Proteins; Cleland, J. L., Langer, R., Eds.; American Chemical Society: Washington, DC, 1994; pp 120-133. (14) Carpenter, J. F.; Chang, B. S. In Biotechnology and and Biopharmaceutical Manufacturing, Processing and PreserVation; Avis, K., Wu, V., Eds.; Interpharm Press: Buffalo Grove, IN, 1996. (15) Randolph, T. W.; Carpenter, J. F. AIChE J. 2007, 53, 1902–1907. (16) Schwegman, J. J.; Hardwick, L. M.; Akers, M. J. Pharm. DeV. Technol. 2005, 10, 151–173. (17) Abdul-Fattah, A. M.; Kalonia, D. S.; Pikal, M. S. J. Pharm. Sci. 2007, 96, 1886–1916. (18) Carpenter, J. F.; Crowe, J. H.; Arakawa, T. J. Dairy Sci. 1990, 73, 3627–3636. (19) Arakawa, T.; Kita, Y.; Carpenter, J. F. Pharm. Res. 1991, 8, 285– 291. (20) Caliskan, G.; Mechtani, D.; Roh, J. H.; Kisliuk, A.; Sokolov, A. P.; Azzam, S.; Cicerone, M. T.; Lin-Gibson, S.; Peral, I. J. Chem. Phys. 2004, 121, 1978–1983. (21) Manning, M. C.; Patel, K.; Borchardy, R. T. Pharm. Res. 1989, 6, 903–918. (22) Ragooonanan, V.; Aksan, A. Biophys. J. 2008, 94, 1–16. (23) Izutsu, K.; Kojima, S. Pharm. Res. 2000, 17, 1316–1322. (24) Heller, M. C.; Carpenter, J. F.; Randolph, T. W. Biotechnol. Bioeng. 1999, 63, 166–174. (25) Suzuki, T.; Franks, F. J. Chem. Soc., Faraday Trans. 1993, 89, 3283–3288. (26) Shalaev, E. Y.; Kanev, A. N. Cryobiology 1994, 31, 374–382. (27) Shalaev, E. Y.; Franks, F. Cryobiology 1996, 33, 11–26. (28) Murase, N.; Franks, F. Biophys. Chem. 1989, 34, 293–300. (29) Heller, M. C.; Carpenter, J. F.; Randolph, T. W. Arch. Biochem. Biophys. 1999, 363, 191–201. (30) Jovanovic, N.; Gerich, A.; Bouchard, A.; Jiskoot, W. Pharm. Res. 2006, 23, 2002–2013. (31) Cavatur, R. K.; Suryanarayanan, R. Pharm. Res. 1998, 15, 194– 199. (32) Remmele, R. L.; Stushnoff, C. Biopolymers 1994, 34, 365–370. (33) Heller, M. C.; Carpenter, J. F.; Randolph, T. W. J. Pharm. Sci. 1996, 85, 1358–1362. (34) Murase, N.; Echlin, P.; Franks, F. Cryobiology 1991, 28, 364–375. (35) Heller, M. C.; Carpenter, J. F.; Randolph, T. W. Biotechnol. Prog. 1997, 13, 590–596. (36) Shouvik, R.; Henderson, I.; Nayar, R.; Randolph, T. W.; Carpenter, J. F. J. Pharm. Sci. 2007, 97, 5132–5146. (37) Fuller, B. J.; Lane, A. N.; Benson, E. E. Life in the Frozen State; CRC Press: Boca Raton, FL, 2004. (38) Pikal-Cleland, K. A.; Carpenter, J. F. J. Pharm. Sci. 2001, 90, 1255– 1268. (39) Gomez, G.; Pikal, M. J.; Rodriguez-Hornedo, N. Pharm. Res. 2001, 18, 90–97. (40) Kahan, T. F.; Reid, J. P.; Donaldson, D. J. J. Phys. Chem. A 2007, 111, 11006–11012. (41) Hare, D. E.; Sorensen, C. M. J. Chem. Phys. 1990, 93, 25–33.

J. Phys. Chem. B, Vol. 113, No. 30, 2009 10087 (42) Luyet, B. J.; Rapatz, G. Biodynamica 1958, 156, 1–68. (43) Abdul-Fattah, A. M.; Kalonia, D. S.; Pikal, M. J. J. Pharm. Sci. 2007, 96, 1886–1916. (44) Carpenter, J. F.; Crowe, J. H. Biochemistry 1989, 28, 3916–3922. (45) Allison, S. D.; Chang, B. S.; Randolph, T. W.; Carpenter, J. F. Arch. Biochem. Biophys. 1999, 223, 289–298. (46) Kaushik, J. K.; Bhat, R. J. Biol. Chem. 2003, 278, 26458–26465. (47) Conrad, P. B.; de Pablo, J. J. J. Phys. Chem. A 1999, 103, 4049– 4055. (48) Aksan, A.; Toner, M. In The Biomedical Engineering Handbook, 3rd ed.; Bronzino, J. D., Ed.; Taylor & Francis: Boca Raton, FL, 2006; Vol. 3. (49) Timasheff, S. N. Biochemistry 2002, 41, 13473–13482. (50) Hubel, A.; Darr, T. B.; Chang, A.; Dantzig, J. Cryobiology 2007, 55, 182–8. (51) Cavatorta, F.; Fontana, M. P.; Vecli, A. J. Chem. Phys. 1976, 65, 3635–3640. (52) Podstawka, E.; Ozaki, Y.; Proniewicz, L. M. Appl. Spectrosc. 2004, 58, 1147–1156. (53) Korber, C. Q. ReV. Biophys. 1988, 21, 229–98. (54) Mullis, A. M. J. Appl. Phys. 1996, 80, 4129–4136. (55) Kurz, W.; Fisher, D. J. Fundamentals of Solidification; Trans Tech: Aedermannsdorf, Switzerland, 1989. (56) Vonnegut, B. J. Appl. Phys. 1947, 18, 593–595. (57) Miller, D. P.; de Pablo, J. J.; Corti, H. Pharm. Res. 1997, 14, 578– 590. (58) Kacˇura´kova´, M.; Mathlouthi, M. Carbohydr. Res. 1996, 284, 145– 157. (59) Malsam, J.; Aksan, A. J. Phys. Chem. B 2009, 113 (19), 6792– 6799. (60) Arakawa, T.; Timasheff, S. N. Biochemistry 1982, 21, 6536–6544. (61) Xie, G.; Timasheff, S. N. Protein Sci. 1997, 6, 211. (62) Webb, S. J. Bound Water in Biological ActiVity; Charles C. Thomas: Springfield, IL, 1965. (63) Porubcan, R. S.; Watters, K. L.; McFarland, J. T. Arch. Biochem. Biophys. 1978, 186, 255–264. (64) Chen, M. C.; Lord, R. C.; Mendelsohn, R. Biochim. Biophys. Acta 1973, 328, 252–260. (65) Perez, C.; Griebenow, K. Biotechnol. Lett. 2000, 22, 1899–1905. (66) Stokkum, I. H. M. v.; Linsdell, H.; Hadden, J. M.; , P. I.; Haris; Chapman, D.; Bloemendal, M. Biochemistry 1995, 34, 10508–10518. (67) Kacura´kova´, M.; Mathlouthi, M. Carbohydr. Res. 1996, 284, 145– 157. (68) Chi, E. Y.; Krishnan, S.; Randolph, T. W.; Carpenter, J. F. Pharm. Res. 2003, 20, 1325–1336. (69) Fonseca, F.; Marin, M.; Morris, G. J. Appl. EnViron. Microbiol. 2006, 72, 6474–6482. (70) Hunter, J. R.; Kilpatrick, P. K.; Carbonell, R. G. J. Colloid Interface Sci. 1990, 137, 462–482. (71) Guzey, D.; McClements, D. J.; Weiss, J. Food Res. Int. 2003, 36, 649–660. (72) Wendorf, J. R.; Radke, C. J.; Blanch, H. W. Biotechnol. Bioeng. 2004, 87, 565–573. (73) Shire, S. J.; Shahrokh, Z.; Liu, J. J. Pharm. Sci. 2004, 93, 1390–1402.

JP809710D