The organic phase consisted of trioctylmethylammonium chloride and nonylphenol pentaethoxylate (Rewopal HV5) as surfactant, octanol as cosurfactant and ...
Eur. J. Biochem. 184,627-633 (1989) 0FEBS 1989
Protein transfer from an aqueous phase into reversed micelles The effect of protein size and charge distribution Ronnie B. G. WOLBERT', Riet HILHORST', Gijsbert VOSKUILEN', Henk NACHTEGAAL ',Matthijs DEKKER', Klaas VAN'T RIET' and Bert H. BIJSTERBOSCHj
'
Department of Biochemistry, Department of Food Science, Food and Bioengineering Group and Department of Physical and Colloid Chemistry, Agricultural University, Wageningen (Received May 22, 1989) - EJB 89 0648
Proteins are spontaneously transferred from an aqueous solution into reversed micelles, provided the aqueous phase has the proper composition. Besides the composition of the aqueous phase, the composition of the organic phase and the properties of the proteins also play a role. We studied uptake profiles of 19 proteins as a function of pH of the aqueous solution. The organic phase consisted of trioctylmethylammonium chloride and nonylphenol pentaethoxylate (Rewopal HV5) as surfactant, octanol as cosurfactant and isooctane as continuous phase. In all cases, except for rubredoxin, proteins were transferred at pH values above their isoelectric point. The pH where maximal solubilization takes place can be described by the relationship: pHoptimum = isoelectric point f0.11 x 1O-j M , -0.97. So, the larger the protein, the more charge is needed to provide the energy required for the adaptation of the micellar size to the protein size. For protein transfer into sodium di-(2-ethylhexyl)sulphosuccinate (AOT) reversed micelles a similar relationship was found. The percentage of protein transferred could be related to the symmetry of charge distribution over the protein. This symmetry was expressed as the YOof random electric moments on a protein that is larger than the effective electric moment of the protein (YOS) [Barlow, D. J. and Thornton, J. M. (1986) Biopolymers 25, 17171. The larger the value of YOS, the more homogeneously the charges are distributed and the lower the percentage transfer. Several fields of application for reversed micelles in biotechnology have been suggested (for reviews see [l -41). One of the potential applications is the extraction of proteins from an aqueous or solid phase via a reversed micellar phase into a second aqueous phase. This double transfer process is based on the ability of reversed micelles to extract proteins from an aqueous solution into their aqueous core. Most proteins retain their native conformation during the transfer process and remain in an active form when incorporated into reversed micelles [1 - 61. Hatton and coworkers have shown that this technique is feasible for the extraction of proteins from a fermentation broth [7] and Leser et al. [8] demonstrated its applicability for the extraction of proteins from the solid phase. Previously [9] we reported that a-amylase could be extracted from an aqueous solution into a reversed micellar medium and re-extracted into a second aqueous phase in a continuous process with a yield of 45% and a concentration factor of eight with respect to enzyme activity. Addition of the nonionic surfactant nonylphenol pentaethoxylate (Rewopal HV5) to the organic phase, led to an increase both in the degree of solubilization
of the enzyme and in the pH range in which solubilization occurs [lo]. By increasing the rate of mass transfer and the distribution coefficient of the enzyme during forward extraction, the yield has been improved to 85% and the concentration factor to 17 [ll]. From the results published to date, it becomes evident that protein partitioning depends on the pH, ionic strength and type of ions present in the aqueous phase, on the type of surfactant, cosurfactant and organic solvent used, on micellar size, on the temperature and on properties of the protein [ 5 , 12- 161. However, no information is available on the mechanisms by which these variables affect protein partitioning. In this study the partitioning behaviour is related to such protein properties as size, isoelectric point and distribution of charged groups over the protein.
MATERIALS AND METHODS Chemicals Trioctylmethylammonium chloride was obtained from Merck, and contained 88% (by mass) of the quaternary ammonium salt, about 10% (by mass) of a mixture of octanol Correspondenceto R. Hilhorst, Department of Biochemistry, P.O. and decanol and 2% (by mass) water. Ethylenediamine, 1,2Box 8128, NL-6700 ET Wageningen, The Netherlands phenylenediamine and isooctane were also from Merck. Abbreviations. AOT, sodium di-(2-ethylhexyl)sulphosuccinate; Nonylphenol pentaethoxylate (Rewopal HV5) was purchased Rewopal HV5, nonylphenol pentaethoxylate; w,, molar water to sur- from Rewo Chemische Werke. Acetyl-p-nitrophenol was from factant ratio. Aldrich, N-succinyl-Ala-Ala-Pro-Phe-p-nitroanilide and MiEnzymes. Alcohol dehydrogenase (EC 1.1.1.1); a-amylase (EC 3.2.1.1); carbonic anhydrase (EC 4.2.1.1); a-chymotrypsin (EC crococcus lysodeikticus were from Sigma. Ultrathin precoated gels (pH 3 - lo), cathode fluid, anode 3.4.21.1); hexokinase (EC 2.7.1.1);lysozyme (EC 3.2.1.17); pepsin (EC 3.4.23.1); peroxidase (EC 1.11.1.7); ribonuclease A (EC fluid, electrode wicks, applicator strips and protein test mix3.1.27.5); superoxide dismutase (EC 1.15.1.1); trypsin (EC 3.4.21.4). ture no. 9 were all from Serva.
628 Insulin (bovine pancreas), cytochrome c (horse heart), was used to determine isoelectric points at temperatures of ribonuclease A (bovine pancreas), trypsin (bovine pancreas), 0 - 12 "C, because the little information that is available in the a-chymotrypsin (bovine pancreas), a-amylase (Bacillus literature suggested that it was permitted to assume that the species), hemoglobin (beef blood), peroxidase (horseradish), pH gradient is temperature independent [20]. bovine serum albumin, lysozyme (chicken egg white) and parvalbumin (rabbit) were from Sigma; trypsin inhibitor (hen egg), carbonic anhydrase (bovine erythrocytes), alcohol de- FPLC hydrogenase (horse liver), hexokinase (yeast) and superoxide The molecular mass of alcohol dehydrogenase was deterdismutase (bovine erythrocytes) from Boehringer; pepsin mined by FPLC (Pharmacia), using a Superose I 2 column (porcine stomach) from Merck, and flavodoxin (Megusphaera and 50 mM potassium phosphate, pH 7.0, or 50 mM ethylelsdetzii strain LC1) and rubredoxin ( M . elsdenii strain LC1) enedianiine pH 13.1 as eluent, at a flow rate of 30 ml/h. were purified as described in [I71 and [18]. Proteins in ammonium sulphate suspension were dialysed Measurements ojenzyme activities prior to use, and protein purity was checked by PAGE. Lysozyme activity was measured as described in [21], carbonic anhydrase activity as described in [22]. Alcohol deProtein pur t it ioning hydrogenase was assayed in 40 mM sodium pyrophosphate. Protein solutions containing 25 pM protein were prepared pH 8.0, containing 20 pM NAD', 18 mM ethanol and 23 pM in 35 rnM ethylenedianiine at the appropriate pH, and the semicarbazide. a-Chymotrypsin activity was measured at A z x owas measured against buffer solution with an Aminco 400 nm with 50 pM N-succinyl-Ala-Ala-Pro-Phe-p-nitroDW-2A spectrophotometer. 2 ml of these solutions were ex- anilide in 0.1 M Tris, pH 8.0. tracted with 2 ml isooctane supplemented with 8 mM triThe assay for peroxidase was carried out in citrate buffer, octylmethylammonium chloride, 2 mM Rewopal HV5 and pH 5.0, with 3.7 mM phenylenediamine and 0.012% HzOl. 0.loh (by vol.) octanol, by rotary inversion of the reaction The reaction was monitored at 492 nm. vials at 120 rpm for 2 min. Phase separation was achieved by centrifugation for 2 min at 2500 rpm in a MSE Super Minor table centrifuge. 1.5 ml of the aqueous phase was separated and the pH and A Z x 0were determined against a blank that RESULTS AND DISCUSSION was prepared similarly. In all cases, the pH after extraction Protein transfer from an aqueous into a reversed micellar was used for the preparation of the uptake profiles. The differ- phase has been shown to depend not only on the composition ences with the pH set before extraction was usually within of both phases but also on the properties of the protein under 0.2 pH units, but could amount to 0.5 pH units at pH values investigation [13, 141. Moreover from studies on the transfer below 5.1.5 ml of the organic phase was subjected to a second of amino acids [23] and proteins [5, 12, 131 into organic solextraction with 1.5 ml 0.5 M potassium phosphate, pH 6.9, vents with the aid of surface active compounds, it became for 30 s with vortexing. Phase separation was achieved clear that electrostatic interactions play an important role. spontaneously. Under these re-extraction conditions, all pro- Solubilization is only observed when the surfactant and proteins are extracted into the second aqueous phase. The protein tein are oppositely charged, and in most cases an increase in concentration in the second aqueous phase was determined ionic strength leads to reduced solubilization [7, 13, 24- 271. by measuring the absorbance at 280 nm against a blank pre- Whereas small proteins are transferred into reversed micelles pared in the same way. Hemoglobin concentrations were de- of the anionic surfactant AOT just below their isoelectric termined at 406 nm and for the determination of the percent- point, for larger proteins the situation is more complicated. age of a-amylase transfer, activity was measured according to Not only are they taken up at pH values further away from the method described by Bernfeld [19]. The percentage of their isoelectric points, but also their uptake profiles are protein transferred was defined as the protein concentration narrower, while in some cases 100% transfer is not achieved in the second aqueous phase with respect to the initial protein 1271. To gain insight into protein properties that determine concentration in the first aqueous phase. whether a protein is taken up and at which pH this takes place, we determined uptake profiles for 19 proteins to reversed micelles of the cationic surfactant trioctylmethylammo/soelectric.facussing nium chloride. The proteins were chosen in such a way that Isoelectric focussing was carried out with Serva ultrathin the M , ranged over 6000 - 100000 and the isoelectric points precoated gels with a pH gradient of 3-10, using an LKB over pH 1 - 11. The uptake profiles for some of these proteins Ultrophor Electrofocussing unit with an LKB 2297 are shown in Fig. 1. Solubilization of the proteins was only Macrodrive 5 power supply. Temperature was controlled found at pH values above the isoelectric point, where the within 0.1 "C (as measured on the gel surface) with an Ultra proteins are negatively charged. In all cases a distinct Kryomat TK 30-D connected to a Julabo P heater for fine maximum was observed, in contrast to the results of Goklen and Hatton for reversed micelles of an anionic surfactant [24], control. 10 p1 of a 4 mg/ml solution of marker proteins no. 9 and which show uptake of small proteins over a wide pH range 4 p1 of a 2 mg/ml solution of the samples were applied in the and bell-shaped curves for larger proteins. As can be seen in middle of the gel. Gels were run for about 60min with a Fig. 1, in some cases a shoulder was observed, the origin of constant power of 3 W. Gels were fixed by immersion in which could not be explained. In Table 1 transfer data for all proteins tested are trichloroacetic acid/water (1:5, by vol.) for 10 min and stained with Coomassie brilliant blue in methanol/acetic acid/water summarized. Of the proteins tested, lysozyme, hexokinase and bovine serum albumin were not transferred to the organic (40 : 10 : 50, by vol.). Using a set of marker proteins of known isoelectric points, phase. For ribonuclease and hemoglobin some precipitation the pH gradient over the gel at 4°C was determined. This line at the interface was observed during the forward extraction.
629 Table 1. Compilation qf data on protein soluhilization in reversed micelles For explanation, see text; PI values were measured at 4°C Protein
10-3x~,
Notation
%S
PI
PHW*
Transfer
P H ~ ~ ~ P I
Yo ADH a-AM BSA CAC a-CH CYC FXN MHB HEX INS LYZ CPV PEP PER RNS RXN SOD PTN PTI
Alcohol dehydrogenase a-Amylase BSA Carbonic anhydrase cc-Chymotrypsin Cytochrome c Flavodoxin Hemoglobin Hexokinase Insulin Lysozyme Parvalbumin Pepsin Peroxidase Ribonuclease A Rubredoxin Superoxide dismutase Trypsin Trypsin inhibitor
68 55 68
58
31
71
25 12.5 15 60 - 64
54 35 80
100 12 12 12 34 40 13.5 6 14.5 25 24.5
97 49 24 74 87 61 29
100
8C PTI
t
FXN
cc
6C z
< cc r
RXN
u Y
a
+
g
10
(L a
Y
PER CAC
2c PEP
L
6
8
10
12
PH
Fig. 1. The percentage of protein transferred from 35 mM ethylenediumine into a solution of 8 mM trioctylmethylammonium chloride und 2 mM Rewopal HVS in isooctane, as a function of the p H OJ the aqueous solution. For explanation of the abbreviations, see Table 1
During the re-extraction of cytohrome c and rubredoxin, small red particles appeared at the interface indicative of some loss of haem. For a-chymotrypsin Luisi et al. [5] observed an optimal transfer at pH 10 - 11. The percentage uptake was dependent on protein concentration and type of organic solvent used. Taking into account that Luisi et al. observed 12% transfer from an unbuffered 20 pM a-chymotrypsin solution into cyclohexane containing 12 mM trioctylmethylammonium chloride, and that the percentage transfer decreases upon replacement of cyclohexane by n-hexane, the 6% we found using a slightly different system and a different technique agrees rather well. Whereas Luisi detected no a-
6.8 5.3 4.9 5.8 8.5 9.9 3.8 5.4-6.5 4.7-5.0 5.9 10.9 5.3 < 1.0 6.9 8.0 3.2 5.3 10.0 4.6
12.9 10.0
8.0 10.5 10.3 4.3 10.5
40 95 0 32 6 55 70 15
6.1 4.7 2.2 2.0 0.4 0.5 5.1 -4.5
0 6.0
5 0
0.1
5.4 4.1 11.0 8.3 3.1 5.9 12.0 9.5
58 8
0.1 3.1 4.1 0.3 -0.1 0.6 2.0 4.9
35 65 56 4 25 70
chymotrypsin activity at all (Luisi, personal communication) we found that some 10% of a-chymotrypsin that was solubilised in reversed micelles retained its activity. This difference is probably due to the fact that the extraction took only 2min, whereas Luisi et al. [5] allowed up to two days for transfer. For a number of other proteins that were solubilized under our experimental conditions, Luisi's group observed hardly and transfer. The experimental details provided do not enable us to explain these differences. Although transfer data for several proteins are described in the literature, only for RNase [28], a-amylase [9 - 1I] and alkaline protease [15] has the effect of transfer on activity been reported. We investigated the effect of transfer on activity for six of the proteins used in this study. For a-amylase and alcohol dehydrogenase hardly any inactivation occurred during the forward and backward extractions. Lysozyme, which was not transferred into the reversed micelles, remained fully active in the first aqueous phase. This enzyme is not inactivated by trioctylmethylammonium chloride, whereas it is by AOT [28]. For a-chymotrypsin, peroxidase and carbonic anhydrase, there was a discrepancy between the amount and activity of protein in the second aqueous phase, indicative of inactivation either during the forward or backward transfer. Further investigations revealed that partial inactivation of peroxidase and a-chymotrypsin occurred during the forward transfer. Carbonic anhydrase on the other hand was inactivated during the backward transfer. Inactivation could be prevented by adaptation of the re-extraction conditions. From Table 1 it can be seen that transfer of small proteins takes place relatively close to the isoelectric point, whereas for large proteins the pH difference is larger. In order to relate quantitatively the pH values of maximum transfer to the isoelectric points, reliable values of isoelectric points are needed. Literature data are not consistent, even when values for proteins from the same source are reported (Table 2). Furthermore, the reported isoelectric points have been determined at temperatures of 4 - 25 "C, using different techniques. This prompted us to determine the isoelectric points as described in Materials and Methods. The results are also shown
630 Table 2. Isoebctric points of proteins used in this study pI values were experimentally determined at 4°C ~
Source
Protein
Experimentally
Literature values
determined
Alcohol dehydrogenase
horse liver
a-Am ylase BSA Carbonic anhydrase
Bacillus sp. bovine serum bovine erythrocytes
a-Chymotrypsin Cytochrome c
bovine pancreas horse heart
Flavodoxin Hemoglobin
M . elsdenii bovine blood
IIexokinase
yeast
Insulin Lysozyme Parvalbumin
bovine pancreas chicken egg white rabbit porcine stomach horseradish
Pepsin Peroxidase Ribonuclease A Ru bredoxin Superoxide dismutase Trypsin Trypsin inhibitor
bovine pancreas M . elsdenii bovine erythrocytes
bovine pancreas hen egg
8.7-9.3 [31] 5.2 [38] 4.7-4.9 1311, 4.97 [35], 4.9 1361 5.3 [30], 6.18 [34], 6.2 [35], 5.9 [36] 8.8 [31], 8.4 [32] 10.65 [36], 9.3 1371, 10.16 1371
6.8 5.3 4.9 5.8 8.5 9.9 3.8 5.4-6.5 4.7 - 5.0
4.5-4.8 [30], P-I 5.3 1311, P-I14.Y [31], 4.7 1311, P-I 5.3 [33], P-I1 5.0 [33] 5.7 [31, 321 10.6- 10.9 [30] 4.9 [33] 2.2-2.8 [31], 3.0-3.2 [33] 7.2 [30], 8.0-8.9 [30], 4.0-8.4 1311, 4.0 [32], 8.8 [32], 6.1 1321, 7.2 [32], 6.5 1331, 7.1 [27] 7.8 [30], 9.3 1311, 7.8 [34], 9.45 [36], 8.9 [37]
5.9 5.3 6.9 8.0 3.2 5.3 10.0 4.6
4.95 [30] 3.8-4.8 1301, 4.8 [30]
in Table2. In most cases the values agree rather well with literature data, but in some cases large differences are observed. For ribonuclease A we obtained a value of 8.0, in good agreement with the value reported in [30] and [31], but differing by an amount of 1.4 pH unit from the value given by Serva [36].Also the value for alcohol dehydrogenase deviates markedly from the value reported by Malamud and Drysdale ~311. Uptake experiments were carried out at 22 "C and isoelectric focussing was originally performed at 4°C. As dissociation constants of amino acids depend on temperature [20, 32, 331, it is relevant to have information on the changes in isoelectric point with temperature. Only for three proteins have dpI/dt values been reported [32, 351. The lower value was -0.009 pH unit/degree, the higher value -0.017 pH unit/degree. Because it is unknown whether these values are extremes, we determined additional isoelectric points for the proteins we used at temperatures ranging over 0- 12°C. Data were interpreted as described in Materials and Methods. The variation of isoelectric points with temperature was found to be 0.00 - 0.03 pH units/degree. As this variation was relatively small, the values of the isoelectric point at 4°C were used to check the dependency of the pH of maximal transfer on the isoelectric point and M,. As already indicated, small proteins are solubilized in reversed micelles close to their isoelectric points, whereas for large proteins the difference between PI and the pH of solubilization increases. This is visualized in Fig. 2, where the difference between the pH of maximum solubilization and the isoelectric point is plotted versus M,. Data for hemoglobin are not incorporated, because both M , and isoelectric point indicate a heterogeneous sample. The line obtained can be described by the equation: PH,,,,-~I = 0.11 x 10-3 M , -0.97.
(1)
The explanation for this observation might be that, as uptake requires adjustment of the reversed micelles to host a protein [39- 421, energy for this process is obtained from
interactions between surfactants and oppositely charged domains of the proteins. The larger the protein, the larger the increase in size of the reversed micelles and the larger the number of charged groups required to compensate the energy expenses. This also explains the shift of the uptake profiles to higher pH,,,-pI values with increasing ionic strength [26]. Because at higher ionic strength the maximal amount of water that can be solubilized decreases, resulting in smaller micelles, and because electrostatic interactions are shielded, a larger number of charged groups on the protein is needed. The intercept of the M , axis might be an indication that proteins of M , less than 9000 can be incorporated without an additional energy requirement from electrostatic interactions. Goklen [27] studied the transfer of cytochrome c, ribonuclease and carbonic anhydrase from a 0.1 M KCl solution into reversed micelles of the cationic surfactant didodecyldimethylammonium bromide in tetrachloroethylene. As compared to our results, transfer profiles are shifted to higher pH values, either because of charge shielding due to the high ionic strength used, or because of precipitation of the protein at the interface. The latter possibility cannot be excluded because only the first aqueous phase was analysed for the disappearance of protein. Goklen [27] studied the transfer of proteins into AOT reversed micelles more extensively. The proteins used could be divided in three groups: firstly, proteins of a size comparable to the micellar size, i.e. of M , about 12000; secondly, proteins of M , about 25000, and thirdly proteins, with M , up to 65000. Small proteins are solubilized completely from a 0.1 M KCl solution over a wide pH range, proteins of intermediate size yield a more bell-shaped profile and large proteins are hardly taken up. To investigate if a relationship between pH,,,-pI and M , similar to our results also exists for this anionic surfactant, pH values where maximum solubilization occurs were determined from their curves. The results are summarized in Table 3. When complete uptake was observed over a wide pH range, the pH value where solubilization is just complete was taken. Because the values for the isoelectric
631
PTI X
I F 3 x M,
Fig. 2. The difference between the p H where maximal solubilization occurs (pH,,,) and the isoelectric point (PI) as a function of M, ,for trioctylmethylammonium chloride reversed micelles. For abbreviations and data, see Table 1 Table 3. Literature data for the solubilization of proteins in 50 mM AOT reversed micelles pH,,, is determined from data in [27]. Column A contains isoelectric points given in [27], column B values measured by us or reported in the literature. Values in column B were used to calculate pHOpt-pI Protein
Source
x M,
PHOPI
Transfer
Isoelectric point A
pHopt-pI
B
YO x -A my 1ase
6 7.8
15 0 100 100
5.9 4.9 9.0 9.5
5.3" 4.9" 8.5" 8.8"
1-Chymotrypsin a-Chymotrypsinogen
bovine pancreas bovine pancreas
55 68 25 25
Cytochrome L Elastase
horse heart porcine pancreas
12.5 26.5
9.6 7
100 65
10.4 8.9
9.Y a 8.5d
L ysozyme Rennin
chicken egg white calf stomach bovine pancreas bovine pancreas bovine pancreas
12 35 13.5 25 26
10.8 3.9 7 6.7 7.3
100 50 100 100 100
11.0 4.9 7.8 10.5 9.3
10.9b 5.2" 8.0" 10.0"
Bacillus
BSA
Ribonuclease Trypsin Trypsinogen ~~
5
0.3 2.5 1.o 1.7' 0.3 1.5 1.9" 0.1 1.3 1.o 3.3 2.0'
~
' Experimentally determined by the authors of this paper. From [30]. From [32]. From [34]. Value obtained with the isoelectric point reported in [27].
points given in [27] deviate markedly from the isoelectric points as determined by us (Table 2) o r from literature data, we used the values compiled in column B of Table 3 for the analysis of the results. Plotting pH,,,-pI vs M , (Fig. 3) shows that pH,,,-pI increases with M,. Three exceptions are manifest: trypsin, rennin, and a-amylase. For a-amylase and rennin, transfer is expected at pH values below 1, assuming that the behaviour can be predicted by the relationship established in Figs 2 and 3. The pH profile obtained might be the result of a combined uptake and denaturation profile, a hypothesis
supported by the fact that a precipitate was observed at the interface [27]. The anomalous behaviour of trypsin cannot be explained. pHOpt-pIvalues for a-chymotrypsinogen and elastase fit to the curve when the isoelectric points given in [27] are used instead of the values in [32] and [34]. Then seven out of ten data points can be fitted to the line: pH,,,-pI = 0.12 x M , - 1.07. It can be concluded that both for an anionic and a cationic surfactant the size of the protein determines the charge density required for its transfer and thereby the p H of uptake. For trioctylmethylammonium
632
3
&
Lz
a a
-
c I
PTN X
~ C H
2l l-
'I
/
40 20
x REN
ADi\
MHB
1
i 40i
0
SYMMETRY OF CHARGE DISTRIBUTION (% 51
uAM
M,
Fig. 3. The difference between the p H where maximal solubilization occurs (pH,,,) and the isoelectric point (PI) as a function of M, for AOT reversed micelles. El, elastase; TPG, trypsinogen; aCTG, cc-chymotrypsinogen; REN, rennin; other abbreviations as in Table 1. Data from Table 3 chloride the slope of Eqn (1) is 0.11, so for each increase of 10000 in M,, uptake takes place about 1 pH unit further away from the isoelectric point. A protein with an MI of 120000 and an isoelectric point of 1, will be transferred at pH 13 when trioctylmethylammoniurn chloride is used as a surfactant. This is the limit of the size of an extractable protein, and explains why hexokinase with an M , of 100000 and isoelectric point around 5, is not transferred. The existence of such a limit is evident from the results reported by Leser et al. [8]. Upon reversed micellar extraction, using 100mM AOT in isooctane with a molar water/surfactant ratio (w,) of 30, of a crude preparation of Escherichia coli proteins of M , up to 60000 were solubilized and when the w, was 8, proteins of M , up to 20000 were extracted. For AOT and trioctylmethylammonium chloride reversed micelles, the intercept of the line differs by an M , of 100. Above, we interpreted the intercept as the maximal M , that can be incorporated into reversed micelles without an increase in micellar size. Sheu et al. [42] demonstrated that the radii of filled and empty reversed AOT micelles, prepared by extraction of cytochrome c from a 0.1 M KCl solution are 4.57 nm and 4.61 nm, respectively. As for both trioctylmethylammonium chloride and AOT a similar intercept is found, one would expect that the size of the reversed micelles is comparable. This is however unlikely, as AOT micelles have, under the experimental conditions used, a M', of about 15 [42], whereas trioctylmethylammonium chloride reversed micelles have a w, of 6 under the experimental conditions used. For trioctylmethylammonium chloride micelles a smaller radius is expected but no data are available. From Fig. 1 it can be inferred that the amount of protein transferred from the aqueous phase into trioctylmethylammonium chloride reversed micelles ranges over almost 0 - 100%. As stated before, charge interactions between surfactant and protein are important, so it is tempting to relate the percentage transfer to the number of charged groups in the proteins. However, no such relationship could be established for the proteins tested. An explanation might be that the charge density is not distributed homogeneously over the surface and can differ between proteins. This hypothesis was checked with the data presented by Barlow and Thornton [43], who calculated local and overall charge densities at pH 7.0 for several proteins on the basis of their crystal structures. They found
Fig. 4. The percentage of protein solubilized in trioctylmethylammonium chloride reversed n7icelles as a function ofthe symmetry of charge distribution over these proteins as given in (361. Abbreviations as in
Table 1
overall surface charge densities to be rather similar. Local charge densities however, showed considerable variation. This variation was denoted % S, i.e. the YOof random electric moments (erroneously called dipole moments) on the protein that had a value larger than the actual electric moment. The value of % S was calculated by randomly distributing the number of charged groups that are present in the protein over a set of surface points and calculating the resulting electric moment. This procedure was repeated 1000 times and the calculated electric moment was compared to the electric moment calculated from the three-dimensional structure of the protein. When % S is high, charges are distributed very even over the protein surface. By plotting the percentage of protein solubilized at pH,,, in the reversed micellar phase against YOS, a good correlation is obtained (Fig. 4). Trypsin inhibitor, which showed non-ideal behaviour in Fig. 2, fits well in this picture, as does the heterogenous hemoglobin. Rubredoxin on the other hand is transferred at a higher percentage than expected. Apparently, proteins having an asymmetric charge distribution are more easily transferred but a protein like lysozyme, which has a symmetric charge distribution, is not transferred at all. When we tried to relate the percentage transfer to the value of the electric moment itself, this turned out not to be a determining factor, probably because it is an overall indicator of the distribution of charges. The data for a-amylase cannot be incorporated in Fig. 4. However, reversing the argument, this figure can be used to predict the YOS for a-amylase. It should be noted, that the data presented here refer to a low surfactant concentration. In general, increasing the surfactant concentration leads to an increase in the percentage transfer. Because Hatton et al. usually work with 50mM surfactant, their uptake percentages are much higher and in many cases reach 100%. For lysozyme, for example, they report 100% solubilization in AOT reversed micelles, whereas no transfer is observed in the trioctylmethylammonium chloride system. This could imply that the distribution of charges over a protein cannot be used to predict their solubilization behaviour in AOT reversed micelles. However, from the work by Steinmann et al. [29] it becomes clear that lysozyme is rapidly inactivated in AOT reversed micelles, but not in cetyltrimethylammonium bromide reversed micelles. The quaternary ammonium surfactant trioctylmethylammonium chloride is not detrimental to lysozyme for no activity loss is observed
633 after placing a lysozyme-containing aqueous phase in contact with a trioctylinethylaininonium chloride reversed micellar solution. The difference in transfer can then be explained by the difference in structure between native and inactivated lysozyme. Data presented in this paper will prove useful for the further development of this liquid/liquid extraction technique. Using Eqn (l), the pH at which a given protein will be solubilized in the organic phase can be predicted. The amount of transfer is determined by the symmetry of charge distribution. In principle, the method presented here allows prediction of the uptake behaviour of a protein when the symmetry of its charge distribution is known. Unfortunately, this is only the case for very few proteins. Application of the knowledge presented here on the separation of a mixture of proteins makes it possible to predict the optimal conditions for extraction. We thank Mrs J. Toppenberg-Fang for typing the manuscript, Mr M. M. Bouwmans for preparing the figures and the Netherlands Technology Foundation (STW) for financially supporting this research project.
REFERENCES I . Luisi, P. L. & L a n e , C. (1986) Trends Biotechnol. 4, 153- 161. 2. Martinek. K., Berezin, I. V., Khmelnitski, Yu. L., Klyachko, N . L. & Levashov, A. V. (1987) Biocutalysis I , 9- 15. 3. Martinek, K., Levashov, A. V., Klyachko, N., Khmelnitski, Yu. L. & Berezin, I. V. (1986) Eur. J . Biochem. 155, 453-468. 4. Hilhorst, R. (1989) in Structure and reactivity in reverse micelles (Pileni, M. P., ed.) Elsevier, Amsterdam, in the press. 5. Luisi. P. L., Bonner, F. J., Pellegrini, A,, Wiget, P. & Wolf, R. (1979) Helv. Chim. Actu 62, 740-753. 6. Martinek, K., Lmashov, A. V., Klyachko, N. L. & Berezin, I. V. (1977) Dokl. .4kad. Nauk SSSR 236,920-923. 7. Woll, J. M., Dillon, A. S., Rahaman, R. S. & Hatton, T. A. (1987) in Protein purification: micro to macro, pp. 11 7 - 130, Alan R. Liss, New York. 8. Lescr, M. E . , Wei, G., Luthi, P., Haering, G., Hochkoepplcr, A., Blochliger, E. & Luisi, P. L. (1987) J . Chim. Phys. 84, 1113 1118. 9. Dekker, M., Van’t Riet, K., Weijers, S. R., Baltussen, J. W. A., Laane, C. & Bijstcrbosch, B. H. (1986) Chem. Eng. J . 33, B27B33. 10. Dekker, M., Baltussen, J. W. A., Van’t Riet, K., Bijsterbosch, B. H., Laane, C. & Hilhorst, R. (1987) in Biocatalysis in organic media (Laane, C., Tramper, J . & Lilly, M. D., cds) pp. 285288, Elsevier, Amstcrdam. 11. Dckker, M., Van’t Riet, K., Bijsterbosch, B. H., Wolbert, R. B. G. & Hilhorst, R. (1989) AIChE J . 35, 321 -324. 12. Meier, P., Imre, E., Fleschar, M. & Luisi, P. L. (1984) in Surfactanfs in solution (Mittal, K. L. & Lindman, B., eds) pp. 9993012, Plenum Press, New York. 13. Goklen, K. E. & Hatton, T. A. (1985) Biotechnol. Prog. I , 6974. -
14. Kadam, K. L. (1986) Enzyme Microh. Technol. 8,266-273. 15. Rahaman, R. S., Chce, J. Y., Cabral, J. M. S. & Hatton, T. A. (1988) Biotechnol. Prog. 4 , 218-224. 16. Dekker, M.. Hilhorst, R. & L a m e , C. (1989) Anal. Biochem. 178, 217-226. 17. Mayhew, S. G. & Massey, V. (1969) J . Biol. Chem. 244, 794802. 18. Gillard, R. D., McKenzie, E. D., Mason, R., Mayhew, S. G., Peel, J. L. & Stangroom, J. E. (1965) Nature 208, 769-771. 19. Bernfeld, P. (1951) Adv. Enzymol. 12, 379-428. 20. Gelsema, W. J. & De Ligny, C. L. (1977) J . Chromatogr. 130, 41 - 50. 21. Absolom, D. R. (1 986) Methods Enzymol. 132, 95 - 180. 22. Pocker, Y. & Stone, J. T. (1967) Biochemistry 6,668-678. 23. Behr, J. P. & Lehn, J. M. (1973) J . Am. Chem. Soc. Y5, 610861 10. 24. Goklen, K . E. & Hatton, T. A. (1986) Proc. ISEC 86, Munich. Vol. 111, 587. 25. Van’l Riet & Dekker, M. (1984) Proc. 3rd E w . Congress Biotechnol., vol. 3, pp. 541 -544, Verlag Chemic, Weinheim. 26. Dekkcr, M., Van’t Riet, K., Baltussen, J. W. A,, Bijstcrbosch, B. H., Hilhorst, R. & Laane, C. (1987) Proc. 4th Eur. Congress Biotechnol., vol. 2, pp. 507 - 51 0, Elsevier Amsterdam. 27. Goklen, K. E. (1986) Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge MA, USA. 28. Meier, P., Imre, E., Fleschar, M. & Luisi, P. L. (1984) in Suyfactanfs in solution (Mittal, K. L. & Lindman, B., eds) pp. 9991012, Plenum Press, New York. 29. Steinmann, B., Jlckle, H. & Luisi, P. L. (1986) Biopolymers 25, 1133 - 1156. 30. Biochemica Information (1975 and 1987) Bochringcr, Mannheim, FRG. 31. Malamud, D. & Drysdale, J. W. (1978) Anal. Biochem. 86, 620647. 32. Righetti, P. G. & Tudor, G. (1981) J . Chromatogr. 220, I1 5 - 194. 33. Righetti, P. G. & Caravaggio, T. (1976) J . Chromatogr. 127. 1 28. 34. Hames, B. D. & Rickwood, D. (1987) Gelelectrop1ioresi.r ofyroteins. A practical approach, IRL Press, Washington DC. 35. Bours, J. (1973) Sci. Tools 20, 29 -34. 36. Serva catalogue (1987) Heidelberg, FRG. 37. Radola, B. J. (1973) Biochim. Biophys. Acta 295, 412-428. 38. Stein, E. A. & Fischer, E. €1. (1960) Biochim. Biophys. Acra 3Y. 281 - 296. 39. Lcvashov, A. V., Khmelnitsky, Yu. L., Klyachko, N. L., Chernyak, V. Ya. & Martinek, K. (1982) J . Colloid Intcvfuw Sci. 88, 444-457. 40. Chatenay, C., Urbach, W., Nicot, C., Vacher, M. & Waks, M. (1987)J. Phys. Chem. 91,2198-2201. 41. Zampieri, G. G., Jackle, H. & Luisi, P. L. (1986) J . Phy.7. Chem. YO, 1849-1853. 42. Sheu, E., Goklen, K . E., Hatton, T. A. & Chen, S.-H. (1986) Biotechnol. Prog. 2, 175 - 186. 43. Barlow, D. J. & Thornton, J. M. (1986) Biopolymers 25, 17171733.
