Solid-liquid equilibrium for proteins in solutions with an unconventional salt (ammonium carbamate): Phase behavior analysis

Fluid Phase Equilibria 443 (2017) 1e8

Contents lists available at ScienceDirect

Fluid Phase Equilibria j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / fl u i d

Solid-liquid equilibrium for proteins in solutions with an unconventional salt (ammonium carbamate): Phase behavior analysis ^ntara Pesso ^ a Filho b, Gisele Atsuko Medeiros Hirata a, c, Pedro de Alca Everson Alves Miranda c, * ~o Paulo, Departamento de Engenharia Química, Rua Sa ~o Nicolau 210, Centro, 09913-030 Diadema, SP, Brazil Universidade Federal de Sa ~o Paulo, Escola Polit ~, 05424-970 Sa ~o Paulo, SP, Brazil Universidade de Sa ecnica, Departamento de Engenharia Química, Caixa Postal 61548, Butanta c Universidade Estadual de Campinas, Faculdade de Engenharia Química, Departamento de Engenharia de Materiais e de Bioprocessos, Av. Albert Einstein ria Zeferino Vaz, 13083-852 Campinas, SP, Brazil 500, Cidade Universita a

b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 September 2016 Received in revised form 24 March 2017 Accepted 31 March 2017 Available online 2 April 2017

In this manuscript, we report the precipitation and crystallization of three proteins (chicken egg-white lysozyme, porcine insulin and bovine insulin) using ammonium carbamate as salting-out agent. For these proteins, data on solubility, metastability limits, and osmotic second virial coefficient (B22) as a function of salt concentration at constant temperature (25.0
C) were obtained. The value of osmotic second virial coefficient (B22) can be regarded as a selection criterion for protein crystallization, as it is related to the interaction potential between protein molecules. Negative values of B22 and large values of the saltingout constant showed that ammonium carbamate is a good precipitating agent for these proteins. Crystallization trials conducted under specific conditions showed that both insulins form amorphous precipitates in ammonium carbamate solutions, which is compatible with the large negative values of B22 measured. Conversely, lysozyme crystals were obtained under all conditions studied, and B22 values for this enzyme were within or close to the crystallization slot. © 2017 Elsevier B.V. All rights reserved.

Keywords: Protein Crystallization Osmotic second virial coefficient Volatile salt

1. Introduction Crystallization and precipitation are unit operations that are widely used in the chemical, pharmaceutical, and food industries. In both operations, a solid phase mainly containing a target compound is formed from a liquid phase. This solid phase formation occurs by shifting liquid phase conditions, such as temperature, pH, and ionic strength. For the precipitation and crystallization of proteins, this solid phase formation is usually achieved by changing the solution pH or by adding another compound such as a salt, a polymer, or an alcohol. Despite their similarity, precipitation and crystallization differ in the type of solid phase formed, amorphous precipitates or crystals, respectively [1,2]. The characterization of the solid-liquid phase behavior of protein solutions may be difficult due to the complexity inherent to protein molecules, and due to the influence of system conditions on the resulting equilibrium (temperature, pH, ionic strength, and the type of solid formed). However, the description of this phase behavior is necessary both to

* Corresponding author. E-mail address: [email protected] (E.A. Miranda). http://dx.doi.org/10.1016/j.fluid.2017.03.031 0378-3812/© 2017 Elsevier B.V. All rights reserved.

design more efficient downstream processes and to develop models to correlate and predict this phase behavior. Two parameters are important to understanding the solid-liquid equilibrium underlying precipitation and crystallization operations. The first one is solubility, which is the concentration of protein in the liquid phase in equilibrium with the solid phase. Protein solubility is a complex function of polar interactions with the aqueous solvent, ionic interactions with salt, and repulsive electrostatic interactions between charged molecules [3,4]. Therefore, this solubility depends on the protein itself and the compound used to reduce the solubility (usually a salt), and its concentration, temperature, and pH. The second important parameter is the osmotic second virial coefficient, B22. The osmotic second virial coefficient is obtained from the Taylor expansion of the osmotic pressure as a function of protein concentration. Its value is a measure of the deviation from the ideality of the solution [5]. The parameter B22 can be associated with the potential of mean force, a measurement of intermolecular interactions between solute molecules in a liquid solution. Large negative values of this parameter indicate a strongly attractive protein-protein interaction, whereas positive values indicate a predominantly repulsive interaction [6].

2

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

The relationship between the B22 value and the outcome of protein precipitation operations was first studied by George and Wilson [7]. Those authors determined B22 values by static light scattering (SLS) for different protein/solvent systems under conditions which did or did not favor the formation of a crystal phase. The formation of crystalline solid phases was observed only in a relatively narrow range of negative B22 values, 8  104e1  104 mL mol/g2. These authors proposed the use of B22 as a tool for predicting conditions under which the formation of amorphous precipitates or crystals occurs. This range was henceforth referred to as the crystallization slot [8]. This hypothesis was also investigated and confirmed by other authors [9e12] using different methods. Velev et al. [11] studied the effects of pH and electrolyte concentration on protein interactions by static light scattering (SLS) and small-angle neutron scattering (SANS). Their results corroborated the hypothesis of the crystallization slot proposed by George and Wilson [7]. Tessier and Lenhoff [10] discussed the relationship between B22 values and solution conditions (pH, ionic strength) that may lead to crystallization of some molecules, such as lysozyme, ribonuclease A, and albumin. According to these authors, there are four different patterns of protein-protein interactions as a function of ionic strength and the type of salt. Liu et al. [12] analyzed and measured experimentally, under different solution conditions, the nucleation and crystallization rates, solubility, crystal morphology, and B22 values to explore phase behavior and lysozyme crystallization. Although the definition of a crystallization slot allows the selection of conditions that favor crystallization, the fact that the B22 value lies within the crystallization slot does not mean that crystals will always be obtained in a precipitation operation [13]. Crystallization depends on both nucleation and crystal growth rates, which depend on the supersaturation, i.e., the difference between the actual protein concentration and the solubility under the same conditions. Thus, knowledge of protein solubility as a function of system conditions (salt concentration, pH, and temperature) is essential to understanding crystallization processes [14,15]. Nevertheless, optimal conditions for the crystallization of proteins are usually determined by trial-and-error procedures [16]. The main reason for this is that systematic studies that present data on B22, protein solubility, and solid phase analysis under the same conditions are scarce. Models relating the B22 and the solubility of proteins have been developed [14,15], but they are not predictive: reliable experimental data are needed to adjust model parameters. This study investigated the solubility, metastability and osmotic second virial coefficient of chicken egg-white lysozyme, porcine insulin, and bovine insulin in aqueous solutions containing ammonium carbamate. This volatile electrolyte dissociates in aqueous solution. The concentration of the ionic species in solution depends on the system temperature and pressure [17]. The electrolyte can be reused without additional purification steps, as it can be converted to the volatile form by changing the system conditions. This salt is an alternative to conventional salts used as precipitant agents, such as ammonium sulfate and sodium chloride [18e20]. To understand the effects of protein-protein interaction, aiming at control of outcome of the solid-liquid equilibrium [21], B22 values for these proteins were determined as a function of the salt concentration. B22 values were determined both by self-interaction chromatography (SIC) and static light scattering (SLS). While SLS can be seen as a conventional method to determine B22, the use of SIC has received increasing attention in recent years as a tool for understanding the phase behavior of proteins [22e27].

2. Experimental 2.1. Materials Porcine (96.6% m/m containing 0.5% m/m of zinc) and bovine (95% m/m containing 0.5% m/m of zinc) insulins were donated by s (Brazil). Chicken egg-white lysozyme (CAS number 12650Biobra 88-33) was obtained from Sigma-Aldrich (USA, catalog number L6876, purity  90%). The proteins were used without further purification. Reagent grade ammonium carbamate (NH4NH2COO, CASRN 1111-78-0) was obtained from Sigma-Aldrich (USA, 99% pure). Ultra pure water was obtained with a Milli-Q system (Millipore, USA). The activated agarose matrix (CNBr-activated Sepharose™ matrix 4 Fast Flow) and Tricorn 5/50 column were from GE Healthcare Life Sciences (Sweden). Hydrochloric acid P.A. ^ (36.5e38.0%) and glacial acetic acid P.A. (99.7%) were from Exodo Científica (Brazil); ammonium bicarbonate (99e100%), from J.T. Baker (Mexico); sodium chloride (grade Ph Eur) and tris(hydroxymethyl)aminomethane (Tris, grade Ph Eur), from Merck (Germany); anhydrous sodium acetate P.A. (99%), from Vetec (Brazil); and ethanol P.A. (99.5%) and toluene P.A. (100%), from Synth (Brazil). All the other reagents were of analytical grade. The purities and sources of these compounds are also presented in Table 1. 2.2. Methods 2.2.1. Solubility measurements Solubility measurements were carried out by dissolution runs conducted in 2.0 mL Eppendorf tubes [28]. Stock solutions of protein and ammonium carbamate had been produced prior to the runs and added to the Eppendorf tubes. The resulting mixture was stirred and placed in a thermostatic bath (model TE-2000 from Tecnal, Piracicaba, Brazil) at 25.0
C. At different times the liquid and solid phases from a tube were separated by centrifugation at 7500g for 15 min. A supernatant phase aliquot was withdrawn with a syringe and filtered (Millex filter, pore size 0.45 mm, Millipore, USA) and its protein concentration determined. Equilibrium was considered to be attained when protein concentration remained constant for 24 h. Porcine insulin and bovine insulin concentrations in solution were determined through the Bradford method [29] using the Coomassie Plus reagent (Pierce, USA). Lysozyme concentration was determined by measuring absorbance at 280 nm. These measurements were made using a DU 640 spectrophotometer (Beckman Instruments, USA). Assays were performed in triplicate and the average protein concentration was considered to be the solubility at the system temperature and salt concentration. 2.2.2. Cloud-point measurements The metastability limit was

estimated

by

cloud-point

Table 1 Sources and purities of compounds used in the experiments. Chemical name

Source

Mass fraction purity

Porcine insulin Bovine Insulin Chicken egg-white lysozyme Ammonium carbamate Hydrochloric acid P.A. Acetic acid Ammonium bicarbonate Sodium chloride Tris(hydroxymethyl)aminomethane Anhydrous sodium acetate P.A. Ethanol Toluene

s Biobra s Biobra Sigma Aldrich Sigma Aldrich ^ Exodo Científica ^ Exodo Científica J.T. Baker Merck Merck Vetec Synth Synth

>96.6 >95.0 90.0 >99.0 P.A. P.A. 99.0e100.0 grade Ph Eur grade Ph Eur P.A. P.A. P.A.

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

measurements. The methodology applied was the same as that used by Watanabe et al. [19]. Electrolyte and protein solutions were filtered through syringe filters (JBR610245 Millex, pore size of 0.22 mm, Millipore, USA) just prior to use. An aqueous salt solution was slowly added dropwise to an aqueous protein solution until the mixture became cloudy. Then the mixture was stirred briefly and stored at 25.0
C in a thermostatic bath and monitored visually. As the solution became clear, more salt solution was added dropwise to the protein solution. This procedure was performed until the mixture had remained cloudy over a period of 24 h. To obtain the cloud-point curve, this experiment was repeated with different concentrations of salt and protein and all assays were performed in triplicate. The mixture composition was calculated as the arithmetic mean of the compositions before and after addition of the solutions. The salt and protein solutions were prepared gravimetrically by dissolving the compounds without pH adjustment. For the lysozyme, protein was weighed and dissolved in deionized water. For the bovine and porcine insulins, proteins were weighed and dissolved in an aqueous low-concentration ammonium carbamate solution (0.06 mol/kg) due to the low solubility of insulin in water. The mass fractions of the protein and ammonium carbamate ranged from 0.02 to 0.20 and from 0.10 to 0.30, respectively. 2.2.3. Determination of B22 using SIC 2.2.3.1. Protein immobilization. Proteins were dissolved in 0.1 mol/l NaHCO3 pH 8.3 at 6.5 mg/mL. A mass of 0.5 g of activated agarose was first washed with a 0.001 mol/L HCl solution at 4.0
C. After washing, 10.0 mL of protein solution were added to the gel. The mixture was left under mild stirring overnight at 4.0
C for the protein to be coupled to the matrix via primary amino groups. After this period, the agarose was washed with 0.1 mol/L NaHCO3 pH 8.3 to remove any unbound protein. The washed agarose particles were added to 15 mL of 0.1 mol/L tris-HCl buffer at pH 8.0, which was used to cap the remaining reactive groups. The mixture was allowed to stand for 2 h. Then the gel was washed five times alternating between the 0.1 mol/L sodium acetate buffer containing NaCl at pH 3.5 and the tris-HCl buffer containing 0.5 mol/L NaCl at pH 8.5 (Instructions-GE Healthcare, USA). The same reaction procedure, but without protein, was applied to another mass of 0.5 g of activated agarose to produce a matrix to determine the retention volume of the column without the interaction between proteins in the mobile phase with the immobilized proteins. The CNBractivated agarose with and without immobilized protein was packed into chromatographic columns. The integrity of the columns was measured by injecting 50 ml of a 1% acetone solution and confirmed by the resulting Gaussian peak. 2.2.3.2. SIC runs. Aqueous solutions of ammonium carbamate were used as mobile phase. Protein solutions at concentrations below saturation in ammonium carbamate solutions were prepared using electrolyte concentrations of 0.06, 0.25, 0.45, 0.69, 0.90, 1.09, and 1.29 mol/kg. The retention volume measurements were performed € using an Akta Purifier FPLC (GE Healthcare, USA) system with a detector at 280 nm. The column was equilibrated with 10 mL of mobile phase at the desired concentration of salt. All analyses were performed at a flow rate of 0.75 mL/min and with 50 mL of sample. The chromatograms were analyzed with 5.11 UNICORN software (GE Healthcare, USA). The same procedure was performed using the column without immobilized protein. 2.2.4. Determination of B22 using SLS The SLS measurements were performed in a Zetasizer Nano ZS (Malvern Instruments, UK). The properties used as input data were also determined: refractive index with a RE40D refractometer

3

(Metler Toledo, USA), density and viscosity of the ammonium carbamate solutions with a DMA 4500 density meter and an AMVn Automated Micro viscosity meter (Anton Paar, Austria), and the increase in refractive index of protein in ammonium carbamate solutions at 25.0
C with a BI-DNDCW differential refractometer (Brookhaven Instruments, USA). For B22 measurements, protein solutions were prepared at each salt concentration. Toluene was used to calibrate the standard light scattering intensity. Saline and protein solutions were analyzed with approximately 1.0 mL of each sample being used in each measurement. For both SIC and SLS measurements, protein solutions were specifically prepared for each experiment and were not used subsequently. 2.2.5. Protein crystallization with volatile salt ammonium carbamate Different concentrations of ammonium carbamate and protein were used in the crystallization tests. These tests were performed at 25
C in a 75 mL glass reactor (HEL group, UK). Temperature and pH were monitored with WinISO E985s-CrystalEyes software (HEL group, UK). For the lysozyme, the protein was weighed and dissolved in deionized water. For the bovine and porcine insulins, the protein was weighed and dissolved in 0.06 mol/kg aqueous ammonium carbamate. The protein solution at the desired concentration was filtered through syringe filters (16534 Minisart, pore size of 0.20 mm, Sartorius, Germany) and added to the reactor with suspended bar magnetic stirring at 120 rpm at 25.0
C. Then solutions with different concentrations of ammonium carbamate were added to the reactor to obtain the desired final salt concentration. The pH and temperature were monitored and remained constant during crystallization. Samples were withdrawn with a syringe every 24 h and filtered and the liquid phase was used to determine protein concentration. The system was considered to have reached equilibrium when the protein concentration had remained constant for 24 h. The solid phase was used for particle morphology characterization by scanning electron microscopy (MEV/EDS: LEO Electron Microscopy, UK). 2.3. Results and discussion 2.3.1. Phase diagram determination The phase diagrams of lysozyme and porcine and bovine insulins were constructed from solubility and metastability limit curves at 25.0
C (Fig. 1). The metastable zone is not as well defined experimentally as the thermodynamic solubility, which is an actual equilibrium condition. Metastable limits were determined by cloud-point for different salt and protein concentrations. The metastable width depends on temperature, the presence of impurities, the cooling rate, and the speed at which metastability is reached (supersaturation rate) [30,31]. In this concentration region, the nucleation rate changes from imperceptible to very fast [32e34]. This can be observed in Fig. 1, which shows that the metastable width is very narrow for the proteins studied here. For the salt concentrations investigated, the lysozyme solubility values were from 2.30 to 7.51 mg/mL, and for the porcine and bovine insulins, values were from 0.80 to 117.01 mg/mL and from 0.28 to 136.60 mg/mL, respectively, at 25.0
C. The behavior that had been found by Bernardo et al. [35] and Hirata et al. [36] was observed during solubility experiments: the amount of protein initially added to a specific ammonium carbamate solution was completely dissolved, and after several days, precipitation occurred. Bernardo et al. [35] postulated that this polymorphic behavior of lysozyme and porcine insulin at 25.0
C and different pH values can be attributed to a change in crystalline structure or degree of crystallinity. Another common feature of these proteins

4

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

Fig. 1. Phase diagrams for lysozyme and porcine and bovine insulins in ammonium carbamate at 25
C. (C) Metastability limit and (B) solubility curves (lines are guides to the eye).

was the salting-out phenomenon: there was a decrease in solubility with an increase in salt concentration. This dependence is consistent with the Cohn [37] equation:

lnðSÞ ¼ b  Κ0 S msalt

(1)

where S is the solubility of protein (g/g of solution); msalt is the salt concentration (mol/kg of solution); K0 S is a constant specific to the salt-protein system (kg/mol of salt); and b is a constant dependent on temperature, pH, and the protein. This equation models the salt's capacity to induce protein precipitation and is the basis for the Hofmeister series: the higher the K0 S value, the more effective is the salt [37]. The experimental data obtained here for protein solubility in volatile salt solution were correlated using the Cohn [37] equation (Table 2). In this work, the K0 S value for lysozyme-ammonium carbamate was higher than that obtained for sodium chloride, the least effective salt for precipitation of lysozyme according to Shih et al. [38], but lower than the values obtained for Na2SO4 and Na2HPO4. However, K0 S values for porcine and bovine insulins were similar (around 3.5), but relatively high, suggesting that ammonium carbamate is more effective in the salting-out for these insulins than the other salts. 2.3.2. Determination of B22 B22 values for lysozyme, porcine insulin, and bovine insulin in aqueous solutions of ammonium carbamate at 25.0
C were determined through SIC and SLS, and the values obtained using these techniques were compared (no B22 data were found in the literature for these proteins in ammonium carbamate solutions). The protein retention volumes in SIC were used to calculate the B22 values (Fig. 2 and Table 3). Increasing the salt concentration had a smaller effect on the retention volume of lysozyme (Fig. 2a) than on those of the porcine and bovine insulins (Fig. 2b and c). The retention volumes of the insulins were higher for higher salt concentrations, which reflects a greater intensity of interaction between the immobilized protein and the protein present in this mobile phase. The range of B22 values measured for lysozyme varied from 20.4  104 to 3.6  104 mol mL/g2 for SLS, while

for SIC they varied from 9.04  104 to 5.12  104 mol mL/g2 (Fig. 3). Considering the experimental uncertainty, these values can be deemed similar, with good agreement between values measured by both techniques except at the highest salt concentration (1.29 mol/kg). For the porcine and bovine insulins, B22 values were more negative with increasing salt concentration, as expected due to the salting-out behavior of these proteins (Fig. 3). The values for B22 measured with SIC and SLS, shown in Fig. 3b and c, were discrepant, despite their showing the same trend: they become more negative when salt concentration increases. The analysis of insulin data is difficult as both insulins form dimers and/or oligomers in solution; insulin molecules form stable hexamers in presence of zinc. This complicates the interpretation of the SLS results [25]. In addition, Doty et al. [42] investigated the monomer-dimer equilibrium of insulin in solution. They concluded that the interactions that result in protein self-association may differ in magnitude, and insulin may exhibit different selfassociation equilibrium constants (e.g., comparing insulin obtained from different sources or samples that underwent slight denaturation in preparation). Alford et al. [43] showed how B22 values measured by light scattering are affected in protein solutions containing monomers and dimers. Even low dimer concentrations contribute significantly to reduce the parameter Kc/R. They developed an expanded coefficient model to measure and quantify the non-ideality of solutions containing monomers and dimers in equilibrium. The molecular mass of porcine and bovine insulins calculated from experimental SLS data are 11.3 ± 0.2 kDa and 11.0 ± 1.2 kDa, respectively. These values indicate that insulin oligomerization does occur, as they are close to the value of the insulin dimer. However, they are insufficient either to affirm that all insulin molecules are dimerized or to deduce which oligomers exist in solution. The experimentally observed values of B22 therefore may be regarded just as apparent values. The corresponding values of R/Kc are presented as a supplementary material. SIC would have the advantage of being insensitive to the presence of impurities or oligomers in solution because UV absorbance is measured: unlike with light scattering, UV absorbance is not influenced by the presence of dimers or impurities. However, it

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

5

Table 2 Cohn [37] equation parameters for proteins in ammonium carbamate solutions at 25
C and literature data for lysozyme in different salts solutions. Parameters

K0 s (kg/mol)

b

Literature

This work

Lysozyme 25
C Shih et al. [38]

Lysozyme 25
C

NaCl pH 9.0

Na2SO4 pH 9.0

Na2HPO4 pH 8.0

NH4COONH2 pH 10.0

0.447 0.903

2.38 3.71

2.095 3.77

1.61 2.45

Porcine insulin 25
C

Bovine insulin 25
C

3.59 4.90

3.51 4.08

Fig. 2. Protein profile in SIC at 25.0
C as a function of ammonium carbamate concentration (in mol/kg): 0.06 (___), 0.25 (…), 0.45 (——), 0.69 (_ _ _), 0.90 (_._), 1.09 (__ __) and 1.29 (__.__). (a) lysozyme, (b) porcine insulin and (c) bovine insulin.

does not mean that oligomerization does not affect elution profiles. A single peak was observed for both porcine and bovine insulins (Fig. 2b and c). The absence of multiple peaks is an evidence (but not a proof) that immobilized protein molecules break the oligomers into their monomeric units, as SLS results clearly show that oligomerization occurs at this conditions. Moreover, size-exclusion chromatograms of proteins exhibiting monomer-dimer equilibrium show distinctive peaks [44]. Insulin chromatograms (Fig. 2b and c) also show large elution volumes. This may be related to the formation of bonds between proteins in immobilized and mobile phases (analogous to those that result in the self-association of insulin molecules in solution), or simply to the fact that the interaction between immobilized and free molecules is stronger than expected for a protein that does not self-associate. Regardless of the mechanism, the observed net effect is a large elution volume, which results in highly negative B22 values for both insulins. Once again, these values of B22 can be regarded as apparent ones. That the values of B22 obtained through SIC are not similar to those obtained through SLS shows that oligomerization affects diversely the results of these techniques. No one of these sets of values can be deemed as strictly correct, and these results point out that additional care must be exercised when measuring the value of B22 for proteins that form stable aggregates. Finally, the highly-negative values of B22 indicate that the stability limit is reached at low protein

concentrations. The stability limit is the condition at which the derivative of the osmotic pressure, as a function of protein concentration, vanishes. For both insulins, the stability limit is close to the protein solubility for each definite salt concentration, but higher than the concentration range investigated in both SLS and SIC experiments. The lysozyme solubility range (from 2.28 to 7.51 mg/mL) as a function of ammonium carbamate concentration was narrow, which indicates little variation in the B22 values for this protein since solubility is in part a result of interactions between protein molecules. The insulins have a wide solubility range in the systems studied (from 0.80 to 117.01 mg/mL and from 0.28 to 136.60 mg/mL for the porcine and bovine insulins, respectively). Despite the high solubility at relatively low salt concentrations, values of B22 were negative for both methods. This indicates a strong attraction between these protein molecules in these ammonium carbamate solutions. A decrease in B22 values as a function of increasing salt concentration is typical of the salting-out behavior (Fig. 3b and c). In the salting-out process, salt ions sequester water molecules and prevent them from forming hydrogen bonds with residues on the protein surface. Consequently, proteins have intermolecular interactions that result in protein-protein attraction [45,46]. Although B22 values for these systems (porcine and bovine insulins in ammonium carbamate) had not been reported in previous

6

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

Table 3 Key parameters for calculating B22 values for lysozyme and porcine and bovine insulins in ammonium carbamate at 25.0
C by SIC. Parameter description

Lysozyme

Porcine insulin

Bovine insulin

Mass protein/particlea (mg/mL) Column void volume/packed particle volumeb Ф - Accessible surface area/mobile phase volumec for protein radiusc (m2/mL) rs e number of protein molecules/accessible surface area (molecules/m2) Ф - Accessible surface area/mobile phase volumec for three protein radiic (m2/mL) BHS - Excluded volumed (cm3)

15.06 0.84

7.66 0.84

37.30 0.84

a b c d

42.5 1.78  10 32.3

43.7 16

6.24  1020

2.17  10 36.2

43.6 16

2.62  1020

10.7  1016 35.8 2.86  1020

Determined experimentally in this work. From DePhillips and Lenhoff [39] for agarose. Interpolated from data for molecules of corresponding size from DePhillips and Lenhoff [36]. Diameters: lysozyme, 3.11 nm; porcine insulin, 2.32 nm; and bovine insulin, 2.39 nm (calculated from molecular volumes [40,41]).

Fig. 3. Comparison of SIC (A) and SLS (◊) experimental data for lysozyme (a) and porcine (b) and bovine (c) insulins as a function of ammonium carbamate concentration at 25.0
C (Lines are the limits of the crystallization slot).

studies, the results are consistent with the literature regarding the general B22 value trend: lower B22 values for higher salt concentrations. However, very negative values such as those obtained in this work suggest the occurrence of precipitation or formation of praggs (aggregates which will result in precipitates [7]). Therefore, the crystallization of porcine and bovine insulins with this volatile salt is not favored at the temperature, pH, and salt concentrations examined in this work. 2.3.3. Protein crystallization with volatile salt ammonium carbamate at 25
C Crystallization runs at 25.0
C were conducted to verify the prediction of the crystallization slot established by George and Wilson [7] (Fig. 4a). For lysozyme, crystals were obtained under all conditions tested. The crystallization trials for porcine insulin at lower ammonium carbamate concentrations from 0.45 to 0.25 mol/ kg resulted in a viscous gel-like solution. Consequently, no crystallization trials were conducted at lower salt concentrations (from 0.45 to 0.06 mol/kg for porcine insulin and from 0.25 to 0.06 mol/kg for bovine insulin). The ranges of B22 values determined for

lysozyme (by SLS from 9.04  104 to 5.12  104 mol mL/g2 and by SIC from 20.4  104 to 3.6  104 mol mL) overlap with the B22 values of the crystallization slot (from -8x104 to -1x104 mol mL/g2), except for the solution with the highest salt concentration. In this range of B22 values, the protein-protein interaction is weakly attractive, favoring the formation of crystals. Another factor contributing to the negative B22 values is the pH of the ammonium carbamate (9.5) being very close to the isoelectric point of lysozyme (pI z 11.0). The electrostatic repulsion between the molecules at pI is minimal. For the porcine and bovine insulins, the formation of crystals was not favored (Fig. 4). According to the literature [47,48], insulin crystals are usually obtained at pH values close to 6.5 in the presence of divalent cations, usually zinc. The cation zinc was not added to the ammonium carbamate solution (pH z 9.5) because at pH values above 8.0 the hexamer (crystalline unit in insulin crystals) is not stable and dissociation of insulin and zinc occurs. Apparently, the molecules of bovine and porcine insulins were agglomerated, in agreement with the very negative B22 values for both proteins: from 6,130  104 to 201  104 mol mL/g2 for porcine and

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

7

Fig. 4. Scanning electron microscope photomicrographs of lysozyme crystals and porcine and bovine insulin precipitates obtained at 25.0
C for different concentrations of ammonium carbamate (in mol/kg): 1.29, 1.09, 0.90, and 0.69.

from 303  104 to 12.9  104 mol mL/g2 for bovine insulin by SIC, and from 250  104 to 18  104 for porcine and from 999  104 to 6.7  104 mol mL/g2 for bovine insulin by SLS. The B22 values fall outside of the crystallization region, except for that for bovine insulin at 0.06 mol/kg ammonium carbamate. These negative values for the B22 parameter indicate the formation of amorphous precipitates. The values of B22 obtained for lysozyme at the highest salt concentration (1.29 mol/kg) are discrepant. We are unable to define unequivocally which value is more reliable, but the limiting value of the molar mass calculated by SLS (10.0 kDa) suggests that this technique may be inaccurate at this condition. This difference may be in part related to the fact that the refractive index increment was obtained at constant salt concentration, and not at constant chemical potential of salt and water [49]. At high salt concentrations, the refractive index increment is higher when measured at constant salt concentration, which results in a lower calculated value of Mw. The observed difference is slightly higher than the expected solely due to this effect [49]. Reliable B22 data could not be obtained at higher salt concentrations due to the constraint imposed by the protein solubility limit. Fig. 4 shows no unequivocal trend relating the average size of crystals and the salt concentration, and crystallization trials were not conducted for salt concentrations higher than 1.29 mol/kg. Should the usual trend of B22 as a function of salt concentration be followed, crystals may be obtained outside the crystallization slot for lysozyme in ammonium carbamate solutions. If it happens, this would confirm that proteins in ammonium carbamate solutions behave atypically. Silva et al. [20] showed that the usual relationship established between the salting-out capacity of a salt and the maintenance of enzymatic

activity is not valid for several enzymes in ammonium carbamate solutions. The value of B22 is related not only to the ability of the salt to crystallize the protein, but also to the influence of the salt on the protein solution behavior. Dumetz et al. [50] studied different proteins in different salt solutions and observed completely different trends of B22 depending on the salt. Therefore, to understand this solid-liquid equilibrium, not only the value of B22 should be analyzed, but also the solubility and the metastable zone width should be assessed. While rare, systematic experiments (with the simultaneous determination of all these properties) are necessary to understand the driving forces of protein precipitation and crystallization. 3. Conclusions The K's values obtained here (from 1.61 to 3.59 kg/mol) were similar to or higher than those for other systems (K's values from 0.447 to 2.38 kg/mol for sodium chloride, sodium sulfate, sodium phosphate, and ammonium sulfate), suggesting a salting-out effect of the volatile salt. Lysozyme crystals were obtained regardless of the B22 value found for different salt concentrations, probably strongly influenced by the pH of the volatile salt solution being very close to the isoelectric point (pI) of this protein. For the bovine and porcine insulins, only amorphous precipitation was favored, which is compatible with the large negative values of B22. The values of B22 correlates with solution conditions that yield protein crystallization or precipitation for lysozyme. However, for proteins that undergo self-association (such as insulin), an expanded coefficient model may be more reliable to quantify the solution non-ideality, as the interpretation of the experimental results is uncertain, and the

8

G.A.M. Hirata et al. / Fluid Phase Equilibria 443 (2017) 1e8

values of B22 obtained by different techniques diverge. While these results establish ammonium carbamate as a good precipitating agent for these proteins and a good crystallization agent for lysozyme, they show that care must be exercised when using the B22 value as a predictor of the outcome of precipitation operations. Acknowledgments The authors gratefully acknowledge the financial support received from the Brazilian agencies CNPq, CAPES, FAEPEX-UNICAMP, and FAPESP (2010/52524-5). The authors thank Prof. Martín Aznar (in memoriam), Prof. Liliane Maria Ferrareso Lona, Prof.  de Almeida Meirelles, and Prof. Watson Loh, Prof. Antonio Jose Eduardo Augusto Caldas Batista for the use of their laboratory equipment. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.fluid.2017.03.031. References [1] A.M. Schwartz, A.S. Myerson, Handbook of industrial crystallization, in: Allan S. Myerson (Ed.), Chap. 1: solutions and Solution Properties, second ed., Butterworth, Woburn, 2002, pp. 1e31. [2] J.M. Prausnitz, Molecular thermodynamics for some applications in biotechnology, Pure Appl. Chem. 75 (2003) 859e873. ^a Filho, On the relationship between the solubility of [3] L.F.M. Franco, P.A. Pesso proteins and the osmotic second virial coefficient, Braz. J. Chem. Eng. 30 (2013) 95e104. [4] C.M. Mehta, E.T. White, J.D. Litster, Correlation of second virial coefficient with solubility for proteins in salt solutions, Am. Inst. Chem. Eng. 28 (2012) 163e170. [5] J.R. Alford, B.S. Kendrick, J.F. Carpenter, T.W. Randolph, Measurement of the second osmotic virial coefficient for protein solutions exhibiting monomerdimer equilibrium, Anal. Biochem. 377 (2) (2008) 128e133. [6] J.M. Mollerup, M.P. Breil, The osmotic second virial coefficient and the GibbsMcMillan-Mayer framework, Fluid Phase Equilibria 286 (2009) 78e84. [7] A. George, W.W. Wilson, Predicting protein crystallization from a dilute solution property, Acta Crystallogr. Sect. D. 50 (1994) 361e365. [8] W.W. Wilson, Light scattering as a diagnostic for protein crystal growth e a practical approach, J. Struct. Biol. 142 (1) (2003) 56e65. [9] B.W. Berger, C.J. Blamey, U.P. Naik, B.J. Bahnson, A.M. Lenhoff, Roles of additives and precipitants in crystallization of calcium- and integrin-binding protein, Cryst. Growth & Des. 5 (2005) 1499e1507. [10] P.M. Tessier, A.M. Lenhoff, Measurements of protein self-association as a guide to crystallization, Curr. Opin. Biotechnol. 14 (2003) 512e516. [11] O.D. Velev, E.W. Kaler, A.M. Lenhoff, Protein interactions in solution characterized by light and neutron scattering: Comparison of lysozyme and chymotrypsinogen, Biophysical J. 75 (1998) 2682e2697. [12] Y. Liu, X. Wang, C.B. Ching, Toward further understanding of lysozyme crystallization: phase diagram, protein-protein interaction, nucleation kinetics and growth kinetics, Cryst. Growth & Des. 10 (2010) 548e558. [13] W.W. Wilson, L.J. DeLucas, Applications of the second virial coefficient: protein crystallization and solubility, Acta Crystallogr. Sect. F. Struct. Biol. Commun. F70 (2014) 543e554. [14] S. Ruppert, S.I. Sandler, A.M. Lenhoff, Correlation between the osmotic second virial coefficient and the solubility of proteins, Biotechnol. Prog. 17 (2001) 182e187. [15] D.F. Rosenbaum, C.F. Zukoski, Protein interactions and crystallization, J. Cryst. Growth 169 (1996) 752e758. [16] R.A. Curtis, J. Ulrich, A. Montaser, J.M. Prausnitz, H.W. Blanch, Protein-protein interactions in concentrated electrolyte solutions e Hofmeister-series effects, Biotechnol. Bioeng. 79 (2002) 367e380. [17] N. Khorshid, MdM. Hossain, M.M. Farid, Precipitation of food protein using high pressure carbon dioxide, J. Food Eng. 79 (4) (2007) 1214e1220. ^a Filho, E.A. Miranda, R.S. Mohamed, Evaluation of [18] E.O. Watanabe, P.A. Pesso the use of volatile electrolyte system produced by ammonia and carbon dioxide in water for the salting-out of proteins: precipitation of porcine trypsin, Biochem. Eng. J. 30 (2006) 124e129. ^ a Filho, Phase [19] E.O. Watanabe, E. Popova, E.A. Miranda, G. Maurer, P.A. Pesso equilibria of lysozyme precipitation with the volatile salt ammonium carbamate, Fluid Phase Equilibria 292 (2010) 42e47. ^a Filho, E.A. Miranda, Evaluation of the effect of ammo[20] L.L. Silva, P.A. Pesso nium carbamate on the stability of proteins, J. Chem. Technol. Biotechnol. 85

(2010) 962e967. [21] Y.-C. Cheng, R.F. Lobo, S.I. Sandler, A.M. Lenhoff, Kinetics and equilibria of lysozyme precipitation and crystallization in concentrated ammonium sulfate solutions, Biotechnol. Bioeng. 94 (1) (2006) 177e188. [22] T. Ahamed, M. Ottens, G.W.K. van Dedem, L.A.M. van der Wielen, Design of self-interaction chromatography as an analytical tool for predicting protein phase behavior, J. Chromatogr. A 1089 (2005) 111e124. [23] T. Ahamed, M. Ottens, B.K. Nfor, G.W.K. van Dedem, L.A.M. van der Wielen, A generalized approach to thermodynamic properties of biomolecules for use in bioseparation process design, Fluid Phase Equilibria 241 (2006) 268e282. [24] R.W. Payne, R. Nayar, R. Tarantino, S. del Terzo, J. Moschera, J. Di, D. Heilman, B. Bray, M.C. Manning, C.S. Henry, Second virial coefficient determination of a therapeutic peptide by self-interaction chromatography, Biopolymers 84 (2006) 527e533. [25] P.M. Tessier, A.M. Lenhoff, S.I. Sandler, Rapid measurement of protein osmotic second virial coefficients by self-interaction chromatography, Biophysical J. 82 (2002) 1620e1631. [26] P.M. Tessier, S.D. Vandrey, B.W. Berger, R. Pazhianur, A.I. Sandler, A.M. Lenhoff, Self-interaction chromatography: a novel screening method for rational protein crystallization, Acta Crystallogr. D. 58 (2002) 1531e1535. [27] D.H. Johnson, A. Parupudi, W.W. Wilson, L.J. Delucas, High-throughput selfinteraction chromatography: applications in protein formulation prediction, Pharm. Res. 26 (2009) 296e305. [28] N. Asherie, Protein crystallization and phase diagrams, Methods 34 (2004) 266e272. [29] M.M. Bradford, A rapid and sensitive method for the quantification of microgram quantities of protein utilizing the principle of protein-dye binding, Anal. Biochem. 72 (1976) 248e254. [30] J. Yu, A. Li, X. Chen, Y. Chen, J. Xie, J. Wu, H. Ying, Experimental determination of metastable zone width, induction period, and primary nucleation kinetics of cytidine 50 -monophosphate disodium salt in an ethanolaqueous mixture, J. Chem. Eng. Data 58 (2013) 1244e1248. [31] K.J. Kim, A. Mersmann, Estimation of metastable zone width in different nucleation processes, Chem. Eng. Sci. 56 (2001) 2315e2324. [32] M. Carpineti, R. Piazza, Metastability and supersaturation limit for lysozyme crystallization, Phys. Chem. Chem. Phys. 6 (2004) 1506e1511. [33] A.M. Schwartz, A.S. Myerson, in: Allan S. Myerson (Ed.), Handbook of Industrial Crystallization, second ed., Woburn, Butterworth, 2002, pp. 1e31. [34] J.M. García-Ruiz, Nucleation of protein crystals, J. Struct. Biol. 142 (2003) 22e31. [35] A. Bernardo, C.E. Calmanovici, E.A. Miranda, Observance of polymorphic behavior during dissolution of insulin and lysozyme, Braz. J. Chem. Eng. 22 (2005) 331e339. [36] G.A.M. Hirata, A. Bernardo, E.A. Miranda, Crystallization of porcine insulin with carbon dioxide as acidifying agent, Powder Technol. 197 (2010) 54e57. [37] E.J. Cohn, The physical chemistry of the proteins, Physiol. Rev. 5 (1925) 349e437. [38] Y.C. Shih, J.M. Prausnitz, H.W. Blanch, Some characteristics of protein precipitation by salts, Biotechnol. Bioeng. 40 (1992) 1155e1164. [39] P. DePhillips, A.M. Lenhoff, Pore size distributions of cation exchange adsorbents determined by inverse size-exclusion chromatography, J. Chromatogr. A 883 (2000) 39e54. [40] M.L. Connolly, The molecular-surface package, J. Mol. Graph. 11 (1993) 139e143. [41] B.L. Neal, A.M. Lenhoff, Excluded-volume contribution to the osmotic 2nd virial-coefficient for proteins, AIChE J. 41 (1995) 1010e1014. [42] P. Doty, M. Gellert, B. Rabinovitch, The association of insulin. I. preliminary investigations, J. Am. Chem. Soc. 74 (8) (1952) 2065e2069. [43] J.R. Alford, B.S. Kendrick, J.F. Carpenter, T.W. Randolph, Measurement of the second osmotic virial coefficient for protein solutions exhibiting monomeredimer equilibrium, Anal. Biochem. 377 (2008) 128e133. [44] H. Bajaj, V.K. Sharma, D.S. Kalonia, A high-throughput method for detection of protein self-association and second virial coefficient using size-exclusion chromatography through simultaneous measurement of concentration and scattered light intensity, Pharm. Res. 24 (2007) 2071e2083. [45] K.D. Collins, Ions from the Hofmeister series and osmolytes: effects on proteins in solution and in the crystallization process, Methods 34 (2004) 300e311. [46] R.A. Curtis, L. Lue, A molecular approach to bioseparations: protein-protein and protein-salt interactions, Chem. Eng. Sci. 61 (2006) 907e923. [47] T. Blundell, G. Dodson, D. Hodgkin, D. Mercola, Insulin: the structure in the crystal and its reflection in chemistry and biology, Adv. Protein Chem. 26 (1972) 279e402. [48] E.J. Dodson, G.G. Dodson, A. Lewitova, M. Sabesan, Zinc-free cubic pig insulin: crystallization and structure determination, J. Mol. Biol. 125 (1978) 387e396. [49] R.A. Curtis, J.M. Prausnitz, H.W. Blanch, Protein-protein and protein-salt interactions in aqueous protein solutions containing concentrate electrolytes, Biotechnol. Bioeng. 57 (1998) 11e21. [50] A.C. Dumetz, A.M. Snellinger-O'brien, E.W. Kaler, A.M. Lenhoff, Patterns of protein-protein interactions in salt solutions and implications for protein crystallization, Protein Sci. 16 (2007) 1867e1877.