Renaturation and identification of periplasmic proteins in two-dimensional gels of Escherichia coli.

The locations of the periplasmic proteins of Escherichia coli on standard two-dimensional gel patterns are described. The periplasmic fractions were prepared by osmotic shock of plasmolyzed whole cells and by release during EDTA-lysozyme treatment of whole cells. Within this fraction of proteins, we identify nine binding proteins (leucine-specific, glutamate-aspartate, glutamine, cystine, galactose, maltose, xylose, ribose, and arabinose) in addition to leucine-isoleucine-valine binding protein, which has been previously identified (Bloch, P. L., Phillips, T. A., and Neidhardt, F. C. (1980) J. Bacteriol. 141, 1409-1420). The identifications are based upon genetic criteria, protein induction, and comigration with purified protein. The levels of these proteins are compared in strains K12, B, and HA12 (a derivative of W). A technique was developed for renaturation of the ligand binding sites of periplasmic binding proteins in denaturing two-dimensional gels. This technique was used to demonstrate that leucine-specific and cystine binding proteins both have different isoelectric points in different strains. Renaturation was also used to demonstrate that there are two different charged forms for glutamine binding protein.

sugar transport. We compare the levels and locations of these binding proteins in representatives of three major E. coli strains. In addition, we describe a method for renaturing the binding sites of periplasmic proteins in denaturing two-dimensional gels. This renaturation technique is used to identify several of the binding proteins in different strains, and should be useful for identification of virtually any protein in which a high aftinity binding site can be renatured.

Strains and Growth Conditions-Wild type E . coli B was obtained
from Professor A. Doermann (University of Washington). Wild type K12 strain was obtained from Professor L. Heppel (Cornell University). Strain HA12-GA7 (an overproducer for glutamate-aspartate binding protein) and its parent strain HA12 were derived previously by several mutations from strain W (5,6).
Except where noted, all cells were grown on the minimal MOPS' medium of Neidhardt et al. (7), supplemented with appropriate carbon sources. Carbon-limited starter cultures were grown overnight with 0.04% (v:v) glycerol. These cultures were then diluted into 10 ml of minimal MOPS medium containing 1% (v:v) glycerol and 1% (w:v) of appropriate sugars for induction. Cells were grown in the latter media at 37 "C with vigorous aeration for at least four generations, and were harvested when the cultures reached an absorbance of 1 to 1.5 at 420 nm (A420). This corresponds roughly to 5 X 1 0 ' cells/ml. Preparation of Periplasmic Fractions-Osmotic shock was performed using a small scale modification of the method of Willis et al. (8). Cultures containing a total of 10 A420 units were preconditioned for shock by the addition of 0.03 volume each of 1 M NaCl and 1 M Tris-HC1 buffer, pH 7.3. Cells were harvested (3000 X g, 10 min, 4 "C) and resuspended in 500 pl of MOPS medium containing I% glycerol and 0.03 volume of 1 M NaCl and 1 M Tris-HC1 buffer, pH 7.3. The resuspended cells were centrifuged for 1.5 min in a Microfuge at room temperature. The resulting pellet was suspended in 50 p1 of 0.03 M Tris-HC1 buffer, pH 7.3. To this suspension were added 50 p1 of 40% (w:v) sucrose in 0.03 M Tris-HC1 buffer, 2 mM EDTA, pH 7.3. The cells were plasmolyzed for 15 min at room temperature, then centrifuged for 1.5 min in the Microfuge. The supernatant was removed. The cells were next mixed vigorously with 100 p1 of cold, distilled water for 30 s. After an additional 15 s on ice, 5 pl of 20 mM MgClz were added, and the suspension was mixed for 10 s. The timing on these steps was critical for a reproducible osmotic shock. The suspension was allowed to stand on ice for several minutes before centrifuging in the Microfuge for 3 min. The supernatant containing the periplasmic proteins was carefully removed and centrifuged again for 3 min in the Microfuge to remove any contaminating cells.
Spheroplasting of cells by EDTA-lysozyme was performed as follows. Cells from 35 A d 2 0 units of culture were harvested (16,000 x g, 10 mi n, 4 "C) and washed once in 10 ml of 30 m~ Tris-HC1 buffer, pH 8.0. The pellet was resuspended in 350 pl of 20% (w:v) sucrose in 30 mM Tris-HC1 buffer, pH 8.0, warmed to 37 "C. A freshly prepared solution of 5.0 mg/ml of lysozyme in 100 mM EDTA was then diluted 10-fold into the suspension of cells at 37 "C. The extent of spheroplasting as a function of time was followed by phase-contrast microscopy and by monitoring susceptability to cell lysis as indicated by The abbreviations used are: MOPS, 4-morpholinepropane sulfonic acid; BP, binding protein.  (13). and cystine binding protein (14) were from K12 sources. Glutamate-aspartate binding protein (15) and glutamine binding protein (16) were from HA12-GA7.  16 and 5e). After electrophoresis, a section of gel (6 x 3 cm) that contained the gel region of interest was cut from the remainder of the two-dimensional gel. This section was transferred to a plastic tissue culture flask of dimensions 6.5 x 3.5 x 2 cm. It was rinsed three times for 10 min each in 10 ml of 50 mM Tris-HCI buffer, pH 7.3, containing 10% (w:v) glycerol (R buffer). It was then incubated for 15 min in 2.5 ml of R buffer containing 2.5 aCi of the appropriate "C-aminoacid or sugar at approximately 3 pM ligand concentration. Subsequently, the gel section was rinsed twice for 10 min each time in 10 ml of R buffer. It was then briefly rinsed in distilled water and dried onto filter paper (37 "C, 2 h) under vacuum.
Kodak X-OMAT AR film was exposed to the dried gel for 75 to 309 h at room temperature.
After autoradiography, the gel was rehydrated in 50% (v:v) methanol and 12% (v:v) acetic acid for 30 min, and the filter paper backing was removed. (Alternatively, some gel sections were immersed in 90% (v:v) methanol and 12% (v:v) acetic acid, which was then gradually diluted to 50% methanol and 12% acetic acid.) The gel section was then swelled in several changes of 10% (v:v) ethanol and 5% (v:v) acetic acid over a 1.5-h period. Fmally it was rinsed three times for 10 min in 10% (v:v) ethanol before silver staining with K the volumes used in the usual fuU size gel procedure. Fig. 1

Proteins
our use of silver stain for visualization instead of autoradiography. Visualization by this particular silver stain requires larger loads of protein than autoradiography, and some distortion in the pattern results. However, the silver stain gives  crisper boundaries for most protein spots, enabling resolution of a number of protein spots that are diffkult to see by autoradiography. Despite these differences in methods, most of the proteins on the patterns of Bloch et al. can be seen to correspond to proteins in our patterns.
In Fig. 2, the locations of the periplasmic proteins are indicated in gray on whole cell position patterns for equilibrium (pH 5-7) and also nonequilibrium isoelectric focusing two-dimensional gels. These are patterns for the locations of whole cell proteins on gels, but they do not contain accurate information about protein spot intensities. The nonequilibrium patterns are included to show the more basic proteins that are not seen on the equilibrium patterns. The grayshaded proteins in Fig. 2 were determined to be periplasmic on the basis of two criteria: 1) release from whole cells by osmotic shock in greater than 50% yield, as judged by comparison between two-dimensional gels of whole cells, shock fluid, and shocked cells (not shown), and 2) release during EDTA-lysozyme spheroplasting of whole cells, as judged from two-dimensional gels of spheroplast supernatant (not shown).
There are a number of other proteins that do not satisfy these criteria, but that are obtained in significant quantities in osmotic shock fluid (cf Fig. 16) and to some extent in spheroplast supernatant. The presence of these proteins appears to result from small amounts of cell lysis. Addition of M$+ earlier in the osmotic shock step did in some cases reduce the amounts of these proteins in shock fluid. On the other hand, the levels of these proteins in shock fluid also appear to be somewhat dependent on growth conditions and treatment of cells before the osmotic shock step.
The equilibrium focusing and nonequilibrium focusing whole cell position patterns and corresponding locations of periplasmic proteins are shown in Fig. 3 for wild type K12 and in Fig. 4 for the W strain HA12. Comparison with the patterns for strain B and the criterion of greater than 50% release during osmotic shock were used to determine which proteins  "Symbols for basis of identification are: CM, comigration with purified protein from the same major strain; R, renaturation of ligand binding site in two-dimensional gels; P, physiological criterion (i.e. induction); G, genetic criterion ( i e . comparison with overproducer mutant); C, protein corresponds to the location of protein identified in another major strain.

pH 5-7 Nonequilibriurn
Leucine-isoleucine- B. K12 " The basis for identification in different strains is given in Table I.
Coordinates for the equilibrium (pH 5-7) gel pattern (Fig. 1B  are periplasmic in the patterns for HA12. For K12, we were unable to get complete release of periplasmic proteins by osmotic shock. The resulting shock fluid was, however, quite free of contamination from cell lysis. We therefore determined the locations of periplasmic proteins for K12 primarily on the bases of the shock fluid pattern and comparison with patterns for B and HA12. The identifications of 10 specific sugar and amino acid binding proteins are shown on the gel patterns in Figs. 2-4. The methods used to identify the binding proteins for the different strains are summarized in Table I. No single strain had detectable levels of all 10 binding proteins. In Table 11, we have given the x and y coordinates that correspond to the locations of these binding proteins on the standard gel patterns published by Bloch et al. (3). These coordinate locations were obtained by comparison between our patterns and those of I

'
FIG. 5. Renaturation of leucine binding proteins in equilibrium (pH 5-7) focusing two-dimensional gels of osmotic shock fluid. a, silver stain of gel section used to renature leucine binding sites for strain HA12 b, autoradiogram of a before silver staining; c, silver stain of gel section used to renature leucine binding sites for strain B; d, autoradiogram of c before silver staining; e, whole gel pattern for osmotic shock fluid from strain B run under nonreducing conditions. Arrows on the left in a-d point to leucine-isoleucinevaline BP; arrows on the right point to leucine BP. Gels contained osmotic shock fluid from 7.6 A4m units of culture grown on glycerol. Electrophoresis and renaturation were performed as described under "Materials and Methods." X-ray film exposures were for 300 h.
Patterns in the gel sections are slightly distorted compared to standard whole cell or shock fluid patterns in which lighter protein loads are used and 2-mercaptoethanol is included. The correspondence between these patterns was determined by comparison between complete two-dimensional gels of osmotic shock fluid run under these different conditions (compare e with Fig. l b ) .
Bloch et al., and were informally confirmed by Prof. Neidhardt's laboratory. 2 Arabinose binding protein and xylose binding protein do not appear in detectable amounts on gels of cells grown on glycerol. We have used arrows to indicate the locations where the principal charged components of these binding proteins would appear on the gel if present. These two proteins are quite similar. They have identical PI values and differ in molecular mass by less than 2000 daltons. Moreover, both proteins frequently appear on gels as a string of three or four charged species, with intensity diminishing in the acid direction. The charge spacing is identical for the two proteins. It appears that this charge heterogeneity develops in vivo, since slab isoelectric focusing of the purified arabinose B P (in the absence of urea) shows the same charge heterogeneity as seen Personal communication by Prof. F. Neidhardt.
for induced xylose B P and purified arabinose BP on the denaturing two-dimensional gels (results not shown). We also note that there is only a small amount of carbamylation of other proteins on our normal gel patterns. Arabinose and xylose BP do not appear to be subject to the same regulation. Xylose BP, but not arabinose BP, was induced when strains B and K12 were grown in the presence of xylose. However, we were unable to induce arabinose B P with arabinose in the presence of glycerol for any of the strains studied here.
Ligand binding to renatured binding sites in the two-dimensional gels was used in the identification of a number of proteins. Fig. 5 shows examples of the autoradiographs and protein stain patterns for sections of gels used to renature leucine binding proteins for strains B and HA12. Also shown is an example of a whole gel run under nonreducing conditions. The major leucine binding component is leucine-isoleucinevaline binding protein. To the right of leucine-isoleucine-valine BP in Fig. 5, b and d, is a less prominent leucine binding component. For strain HA12 (Fig. 5, a and b), this protein corresponds to the location of leucine-specific binding protein identified in K12. For strain B (Fig. 5, c and d), the minor leucine binding component has a different charge. We conclude that the PI for leucine-specific BP in strain B is different from that for HA12 and K12. A similar type of charge difference was also encountered for cystine binding protein in strains HA12 and K12, as can be seen in the identifications in Figs. 3 and 4. Purified glutamine binding protein comigrated with two periplasmic proteins of different charges in strain HA12. Renaturation experiments demonstrated that both of these proteins bind glutamine (results not shown). The two different charged forms do not appear to be caused by artifactual degradation in the two-dimensional gels, since slab isoelectric focusing of purified protein (in the absence of urea) showed the same two spots, and their relative intensities are comparable to the relative intensities of the corresponding proteins in two-dimensional gels of whole cells. The same two periplasmic protein spots appear in strains B and K12 in roughly comparable amounts.
Several of the binding proteins listed in Table I are identified in the gel patterns for one or more strains on the basis of correspondence with the location of the same protein identified on gels of another strain. Proteins identified in this way are listed in parentheses on the gel patterns of Figs. 2-4. Because more than one protein occasionally appears in the same location on gels, it seems wise to treat such correspondences as a tentative guide for the locations of binding proteins until other methods establish the identifications more f d y .

DISCUSSION
The classification of periplasmic proteins on whole cell twodimensional gel patterns for E. coli represents the fmt step in a comprehensive program to classify all whole cell proteins on two-dimensional gels according to subcellular location. Classification of different subcellular fractions serves to extend the general information about physiology and genetic composition that can be obtained from two-dimensional gel electrophoresis of whole cells. Moreover, classification can be used to eliminate some of the ambiguities that can arise in the identification of specific proteins on two-dimensional gel patterns. For example, induction generally causes the elevation of a number of different proteins on two-dimensional gel patterns. It is nevertheless possible to identify many of these proteins if their subcellular classifications are known.
A number of dificulties can be encountered in classifying whole cell proteins on the basis of subcellular location. Distinctions between different subcellular classes are sometimes vague, and some proteins can be located in more than one class simultaneously or can be located in different classes under different conditions. Furthermore, it is often difficult to establish the validity of operational methods for isolating proteins of any given class. In the work described here, we have attempted to rely on a definition of periplasmic proteins as proteins that are freely soluble in the space between the inner and outer membranes of Gram-negative organisms. We have used osmotic shock and EDTA-lysozyme treatment as methods for releasing periplasmic proteins from cells. Osmotic shock is a convenient method that is frequently used for preparing periplasmic proteins. It generally provides nearly quantitative release of periplasmic proteins. However, it is often accompanied by some cell lysis, and it is a mechanically disruptive method which probably releases some proteins that are more properly considered extrinsic inner or outer membrane proteins. EDTA-lysozyme treatment of cells is a more gentle method for releasing periplasmic proteins from cells, but it often gives poor yields and may cause some structural disruption. In designating the periplasmic proteins on whole cell two-dimensional patterns, we have tried to take a conservative approach. Only those proteins that are released from cells in greater than 50% yield by quantitative osmotic shock or that are released in proportionately high yield by EDTAlysozyme treatment or mild osmotic shock are designated as periplasmic proteins. We have thus eliminated most proteins that appear in periplasmic preparations as a result of cell lysis or partial release from the membranes. On the other hand, proteins that are located in both the periplasm and cytoplasm or membranes are probably not identified as periplasmic proteins by these criteria. For each of the three strains studied here, we fiid approximately 50 periplasmic proteins out of about 800 total whole cell proteins on the two-dimensional gel patterns. In general there is good agreement between the periplasmic pattern assignments for the different strains. Most of the differences in the patterns result from variations in proteins between strains. Only in a few cases are corresponding proteins designated as periplasmic proteins in one strain but not another. It is tempting to speculate that these few apparent discrepancies may actually reflect architectural differences between strains. In this respect, it is interesting that we were never able to get quantitative release of periplasmic proteins from K12 by osmotic shock, whereas we had no difficulty with the other strains.
There are a number of interesting features in the binding proteins identified here on two-dimensional gels. Arabinose binding protein and xylose binding protein have extremely similar electrophoretic properties (size, PI, and charge heterogeneity). The degree of similarity between these two proteins is much greater than for any other binding proteins identified. On this basis we predict that arabinose and xylose BP have much greater sequence homology than other sugar binding proteins (ribose, galactose, etc.). Two binding proteins, leucine-specific BP and cystine BP, have different PI values in different strains. There is some possibility that the difference in PI observed for cystine BP in K12 and HA12 is due to mutctions that occurred during construction of strain D2W (parent for HA12) from W. These mutations were known to involve a gene that affects some aspect of cystine transport (5). That gene may code for cystine BP. It would therefore be interesting to compare the gel location of cystine BP for strains HA12 and wild type W. Three of the binding proteins, xylose BP, arabinose BP, and glutamine BP show more than one charged form on the two-dimensional gels. In the case of xylose BP and arabinose BP, this is clearly the result of protein alteration, as evidenced by the charge profiie that consists of several equally spaced species with intensity de-creasing in the acid direction. Apparently this alteration occurs in vivo. It is less clear whether protein alteration has occurred for glutamine BP. Only two glutamine binding components are found, and their relative intensities are quite reproducible. Corresponding proteins are seen in strains B and K12. It seems unlikely that there are two genes for glutamine BP, although the possibility cannot be completely discounted.
The procedure that we have developed for renaturation of ligand binding sites in two-dimensional gels offers possibilities for a number of different experiments. It w i l l be interesting to see whether renaturation in denaturing two-dimensional gels works for other types of proteins besides periplasmic binding proteins. Enzymes and membrane receptors are of particular interest, and the procedure could be very valuable for identification of drug and hormone receptors. Histones and several enzymes have already been renatured from one-dimensional sodium dodecyl sulfate gels by other workers (19)(20)(21)(22)(23)(24)(25).
Several general comments about the renaturation procedure used here are worthwhile. In order to detect bound ligand above the background ligand concentration in the gel, mass action predicts that the concentration of binding sites in a gel spot must be greater than the effective dissociation constant for ligand. We estimate that the maximum protein concentration in a sodium dodecyl sulfate-gel spot is in the range 1-10 p~. Therefore the effective KD for ligand must be micromolar or less if binding is to be easily detected in proteins with single binding sites. Periplasmic binding proteins typically have affinities for ligands that require greater than micromolar concentrations of ligand for saturation (26). Once ligand is bound to these proteins, however, the rate of ligand release is quite slow, and so the effective KO is much less than micromolar. This makes it possible to rinse the gels to remove unbound ligand, thereby reducing background. We have found that renaturation of ligand binding sites in periplasmic binding proteins is ineffective if the proteins have been exposed to sulfhydryl reducing agents. For other proteins, sulfhydryl reduction may not necessarily interfere with renaturation in two-dimensional gels. It is worth noting that several of the proteins renatured from one-dimensional sodium dodecyl sulfate-gels by other workers (19)(20)(21)(22)(23)(24)(25) had been treated with sulfhydryl reducers. We are presently pursuing the possibility of renaturing other types of proteins in two-dimensional gels.