- Email: [email protected]
Is Nitrocellulose Filter Binding Really a Universal Assay for Protein–DNA Interactions?
Analytical Biochemistry 268, 330 –336 (1999) Article ID abio.1998.3056, available online at http://www.idealibrary.com on
Is Nitrocellulose Filter Binding Really a Universal Assay for Protein–DNA Interactions? Stefan Oehler, 1 Regina Alex, 2 and Andrew Barker 3 Institut fu¨r Genetik der Universita¨t zu Ko¨ln, Weyertal 121, 50931 Ko¨ln, Germany
Received September 24, 1998
The ability to bind to nitrocellulose is commonly accepted as being a universal property of proteins and has been widely used in many different fields of study. This property was first exploited in the study of DNAbinding proteins 30 years ago, in studies involving DNA binding by the lactose repressor (LacR) of Escherichia coli. Termed the filter-binding assay, it remains the quickest and easiest assay available for the study of protein–DNA interactions. However, the exact mechanism by which proteins bind to nitrocellulose remains uncertain. Given the supposedly universal nature of the interaction, we were surprised to notice that certain LacR variants were completely unable to bind simultaneously to DNA containing a single lac operator and nitrocellulose. Investigation of this loss of binding suggests that LacR requires a protein region that is both hydrophobic in nature and more or less unstructured, in order to bind to both nitrocellulose and DNA. In the case of wild-type, tetrameric LacR, the DNA-recognition domain that is not bound to DNA suffices. Dimeric LacR variants will only bind if they have certain C-terminal extensions. These experiments sound a cautionary note for the use of filter binding as an assay of choice, particularly in applications involving screening for the DNA-binding site of putative DNA-binding proteins. © 1999 Academic Press Key Words: filter-binding assay; nitrocellulose binding; mobility shift assay; protein–DNA recognition; Lac repressor.
The ability of proteins to bind to nitrocellulose has been recognized since at least the 1930s, when the 1 Present addresses: Department of Genetics, University of Cambridge, Downing St., Cambridge, CB2 3EH, England. 2 Present address: AMGEN GmbH, Reesstr. 25, 80992 Mu¨nchen, Germany. 3 To whom correspondence should be addressed. Fax: 149 221 470 5170. E-mail: [email protected].
330
necessity of using serum to prevent adsorption of virus particles to collodion (cellulose nitrate) membranes during filtration was described (1). In the 1960s, the discovery that ribosomes (2) and an RNA polymerase– DNA complex (3) bound to nitrocellulose filters rapidly led to the application of the property to other systems, including studies involving the DNA-binding of transcription factors (4, 5). The universal applicability of protein adsorption to nitrocellulose is perhaps best underlined by a relatively recent application, but one which is the most widely used: the electrophoretic transfer of denatured proteins to nitrocellulose for subsequent immunological characterization, commonly referred to as Western blotting (6). The exact mechanism, however, by which proteins bind to nitrocellulose remains unclear. Nitrocellulose is a fully nitrated derivative of cellulose, in which all free hydroxyl groups have been substituted by nitrate groups, and is thus hydrophobic in character. Nitrocellulose filters are structured rather like a sponge, through which individual pores follow tortuously winding paths and constitute about 80% of the total filter volume (7). It has long since been apparent that the process of protein binding involves active adsorption, as interacting particles such as viruses or ribosomes are clearly much smaller than the pores of the filter to which they were binding (2, 7, 8). Consideration of the conditions under which adsorption is optimal, and the chemical nature of nitrocellulose, led to the conclusion that binding requires hydrophobic interactions between protein and nitrocellulose, so it is not surprising that some proteins, notably those of small molecular weight, bind with reduced affinity (8 –10). As noted above, the nitrocellulose filter-binding assay has been widely used in analyzing both the thermodynamics and kinetics of protein–DNA interactions (5, 10 –15). Other assays have since become available. One of these is based on protection against base methylation or DNaseI cleavage by a bound protein, which 0003-2697/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved.
IS NITROCELLULOSE BINDING UNIVERSAL?
may be described under the generic name of footprinting (16 –18). The application of magnetic beads (19, 20) or of quenched-flow techniques (21) has enabled the handling times to be sufficiently shortened for the DNase protection assay to become amenable for kinetic studies. The near ubiquitous band shift assay (22, 23) has also proved applicable, at least with short target DNAs. The increased stability of protein–DNA complexes, once they have entered the gel matrix, means that this type of assay is also suitable for determination of kinetic dissociation rates (24). More recently, methods based on fluorescence anisotropy (25) or on plasmon resonance technology (26) have also appeared. These assays all have their strengths and weaknesses, a comparison of which is beyond the scope of this article. However, it may safely be said that the biggest advantages of filter binding in comparison to all other available methods are the ease and rapidity with which a large number of samples may be analyzed, expensive equipment is not required, and it is the assay of choice when analyzing protein binding to operator sequences contained in DNA molecules that are greater than a few kilobases in length. Previously, we reported the difference in association rate with operator located on a l DNA molecule between wild-type, tetrameric LacR, and a dimeric LacR variant, LacR adi (27, 28). We wished to examine this phenomenon further, with additional LacR variants that are also dimeric (29 –31). However, some of these proteins do not bind consecutively to DNA and nitrocellulose. We have undertaken a detailed examination of this loss of binding and present the results, and their implications for the study of protein–DNA interactions generally, here. MATERIALS AND METHODS
Protein and DNA Preparation LacR variants were purified as described previously (28, 32), with one exception. The LacRBasic variant exhibits a much tighter binding to phosphocellulose than other variants (data not shown) and therefore requires elution with a KPO 4 gradient of 0.075–1 M KPO 4, instead of a 0.075– 0.4 M KPO 4 gradient, as in the standard protocol (28). Otherwise, all steps in the purification were identical. Target DNAs were plasmids pEE4 (nonoperator) and p310 (containing O id), or, in the case of band shift assays, the 60- or 84-bp EcoRI–HindIII fragment of each plasmid, respectively. The plasmids, and methods used in purifying plasmid DNA, were as described previously (27, 28). For labeling, plasmid DNA was linearized with EcoRI, or with EcoRI–HindIII, and DNA labeled with [a- 32P]dATP (Amersham), by fill-in reaction using DNA Polymerase I Klenow fragment (Boehringer-Mannheim). Full-length plasmid DNA was then
331
separated from unincorporated label by passing over a Nick-column (Pharmacia), while the EcoRI–HindIII fragment was purified by electrophoresis on 6% nondenaturing acrylamide gels. Filter-Binding and Band Shift Assays In our standard assay (27), filter binding is performed with nitrocellulose disks 25 mm in diameter and with a mean pore size of 0.45 mm. Filters were pretreated by boiling for at least 5 min in distilled water and then, after washing in cold distilled water, allowed to equilibrate for at least 30 min in FB [10 mM Tris–HCl, pH 7.2, 0.1 mM EDTA, 3 mM Mg acetate, 10 mM KCl, 5% (v/v) dimethyl sulfoxide]. Binding reactions were performed in a final volume of 100 ml, in BB (FB containing in addition 0.1 mM dithiothreitol and 50mg/ml bovine serum albumin). Samples were allowed to equilibrate for 30 min before being filtered through a single filter disk (prewashed with 100 ml FB) with a filtering time of 5–10 s and washed with an additional 400 ml FB. Filters were allowed to air dry before being dissolved in 5 ml scintillant (Zinsser Analytik Quickszint 361) and counting. Reactions were performed with a DNA concentration of 5 3 10 211 M; protein concentration, calculated per dimer, was as indicated. Binding reactions for band shift assays were performed under identical conditions as for filter binding, except that the final reaction volume was 10 ml. Samples were mixed with 4 ml tracker dye [BB containing 0.04% (w/v) each bromophenol blue and xylene cyanol and 15% (w/v) Ficoll] and immediately loaded onto 6% polyacrylamide gels. Running buffer was 0.53 TBE (23) and gels were prerun at 20 V/cm until the initial resistance had decreased by at least half. Electrophoresis was performed for 2 h at 10 V/cm and at room temperature. After electrophoresis, gels were dried on Whatman 3M paper and exposed to X-ray film for detection of free and retarded DNA bands. Filtration of samples for analysis in band shift assays was performed as follows. Nitrocellulose disks of suitable size were pretreated as described above and then placed between a 5-ml syringe and the plastic base of a 19-gauge needle, from which the needle had first been removed. Samples in a total volume of 20 ml were pipetted directly on to the filter and then pushed through with the plunger at sufficient pressure to give a filtering time of about 1 s. In the case of protein–DNA mixtures, 10 ml of the filtrate was then processed for loading onto gels as described above. In the case of protein mixtures, samples were filtered at 23 final concentration and then used in binding reactions as described above. For reactions where samples were to be simultaneously analyzed by band shift and filter binding, the reaction was performed in a total volume of 110 ml. First, 10 ml was removed and loaded on
332
OEHLER, ALEX, AND BARKER
The results of a typical filter-binding experiment, examining the binding at equilibrium of these LacR variants to an 84-bp DNA fragment containing the ideal lac operator, O id, are shown in Table 1. A subset of dimeric LacR variants is no longer capable of recording a positive result in this test. Specifically, those variants in which the C-terminal extension is completely removed, or in which residues downstream of position 330 are replaced with the C-terminal 13 amino acids of GalR, appear to have lost the ability to bind to DNA, as determined by filter binding. Additionally, a LacR variant with an extremely basic extreme Cterminus only shows residual binding in this test (Table 1). The Apparent Loss of Binding Is a Loss of the Ability to Bind Simultaneously to Nitrocellulose and DNA
FIG. 1. Amino acid sequence of the extreme C-termini of the LacR variants used in this study: (a) wild-type, tetrameric LacR (51, 52); (b) LacRL 346A (29); (c) LacR adi (35); (d) LacR 331Stop (30); (e) LacR 331Gal (31); and (f) LacRBasic (31). The boxed residues in the sequence of wild-type LacR are those that comprise the leucine heptad repeat (29). Those residues that differ from wild-type LacR are underlined in the other sequences. Note that the substitution L 330S is neutral in wild-type LacR and, where it occurs, was introduced purely for convenience in DNA manipulation (30). The first residue shown in all cases is V 321.
polyacrylamide gels as described above, before the remainder was filtered through nitrocellulose. RESULTS
Apparent Loss of DNA Binding in Filter-Binding Experiments LacR consists of three structural domains: an Nterminal extension (amino acids 1– 61) that comprises the DNA-binding domain; a globular core (amino acids 62–330) that comprises the bulk of the protein, which contains the inducer-binding domain and the dimerization domain between monomers; and a C-terminal domain (amino acids 331–360) that includes two leucine heptad repeats which allow two dimers to aggregate to a tetramer (29, 33, 34). Mutations which disrupt the heptad repeat lead to dimeric forms of LacR, which retain DNA-binding activity but have a dramatically reduced ability to bind cooperatively to two operators (29 –31, 35–38). The amino acid sequences of the extreme C-termini of wild-type, tetrameric LacR and a number of dimeric LacR variants are shown in Fig. 1. As all variants are identical in both DNA-binding and core domains, they should be unaltered in binding affinity, at least to a short DNA fragment carrying a single lac operator.
The same binding mixes assayed by filter binding in Table 1 were also examined in band shift assays (Fig. 2). It is immediately apparent that all LacR variants examined are capable of binding to the DNA fragment in this experiment. Thus, the result presented in Table 1 does not reflect an inability of the LacR variants in question to bind DNA, but an inability to bind simultaneously to DNA and to nitrocellulose. This can also be demonstrated by passing the binding reactions through a nitrocellulose filter before loading on an acrylamide gel. The retarded band (but not the free DNA band) is removed if the binding reaction is filtered before loading for wild-type LacR, LacRL 346A, and LacR adi, but not for LacR 331Stop, LacR 331Gal, or LacRBasic (Fig. 2). A detailed examination of this phenomenon is presented in Fig. 3, where two dimeric LacR variants, LacR adi and LacR 331Stop, are shown as representative nitrocellulose-binding and nonbinding variants, respectively. Both variants were titrated against a fixed
TABLE 1
DNA-Binding Activity of LacR Variants as Determined in Filter-Binding Assays LacR variant a
% input DNA retained on nitrocellulose filters b
None Wild-type LacR LacRL 346A LacR adi LacR 331Stop LacR 331Gal LacRBasic
0.5 6 0.3 24 6 0.9 48 6 2.6 48 6 4.4 0.7 6 0.3 1.2 6 0.6 7.8 6 0.4
a Purified LacR was added at a 50-fold molar excess, calculated per LacR dimer. Target DNA is identical to that used in Fig. 2. b Retention of 5% or less of the input target DNA on nitrocellulose filters is scored as nonbinding (5, 11, 27).
IS NITROCELLULOSE BINDING UNIVERSAL?
FIG. 2. All LacR variants bind DNA but not all bind DNA and nitrocellulose consecutively. Binding reactions, with a 50-fold molar excess of protein, calculated per dimer, were analyzed in a band shift assay. The LacR variants examined are indicated above the figure. In each pair of reactions, the lane on the left was loaded directly after incubation, while the lane on the right was subjected to filtration through nitrocellulose immediately before loading. This is indicated by the presence of a “2” or a “1,” respectively, above each lane and the word “filtered” to the left of the figure. The positions of the slot, protein–DNA complexes, and the free DNA band are also indicated.
333
We undertook additional control experiments to ensure that the lack of binding we observe is not due to some special property of the particular type (Sartorius, mean pore size 0.45 mm) of nitrocellulose filter used routinely in this laboratory. Again, LacR 331Stop and LacR adi were used as representative nonbinding and binding LacR variants, respectively. Identical results to those presented above were obtained with nitrocellulose of a comparable mean pore size manufactured by Millipore, Amersham, Schleicher & Schuell, and Gelman Sciences and Sartorius nitrocellulose filters with a mean pore size of 0.22 mm or other membrane types also available for protein immobilization (data not shown). The results presented here are reminiscent of those obtained with the Escherichia coli Trp repressor
concentration of target DNA. In band shift assays, binding is observed when an excess of either LacR adi or LacR 331Stop is added (Fig. 3a), but again, saturation binding of the operator fragment is only observed for LacR adi if the samples are analyzed by filter binding (Fig. 3b). Thus, the lack of nitrocellulose binding observed for LacR 331Stop is independent of the amount of protein added. LacR Needs a Disordered Domain for Nitrocellulose Binding A simple explanation for the inability of certain LacR variants to bind simultaneously to nitrocellulose and DNA is that both are bound by the same protein domain, namely the DNA-binding headpiece. To test this idea, we performed a band shift assay using the same LacR variants shown in Fig. 2, in which the protein was either untreated or filtered through nitrocellulose filters prior to addition to DNA. The result, shown in Fig. 4, clearly demonstrates that all variants tested bind to nitrocellulose. After preincubation with DNA, some dimeric LacR variants fail to bind to nitrocellulose (Table 1, Fig. 2, Fig. 3). Thus, in these cases, the only protein domain capable of nitrocellulose binding is the headpiece. Those LacR variants still capable of binding simultaneously to nitrocellulose and operator DNA are those in which there is more than one such domain—the second, non-DNA-bound headpiece in the case of wild-type, tetrameric LacR or the (presumably) disordered C-terminal extension in the case of LacRL 346A and LacR adi.
FIG. 3. The lack of nitrocellulose binding by LacR 331Stop is not dependent on protein concentration. (a) Band shift assay performed with LacR adi and LacR 331Stop. The LacR variants used are indicated above the gel, and increasing protein concentration is indicated by a wedge. Lanes in which no protein was added are indicated by a “(2)” sign. The positions of protein–DNA complexes and free DNA are indicated to the left of the gel. (b) Filter-binding assay performed with the same samples. Data points for LacR adi are indicated by open circles, while data points for LacR 331Stop are indicated by filled squares. Protein concentration, calculated per dimer, is indicated. In the absence of protein, DNA retention on the filters was 5 6 0.5% of the input DNA in this experiment.
334
OEHLER, ALEX, AND BARKER
binding must be due to the basic tail of LacRBasic, with the non-DNA-bound headpieces left available for nitrocellulose binding. DISCUSSION
The Significance of the LacR–Nitrocellulose Interaction We have shown that the reason why certain LacR variants cannot bind simultaneously to DNA and nitrocellulose is that the presence of at least one freely accessible structural domain, protruding from the glob-
FIG. 4. LacR alone is capable of binding to nitrocellulose. A band shift assay is shown in which the indicated LacR variants were either mixed directly with target DNA or passed through nitrocellulose filters immediately prior to mixing with target DNA. Protein was added at a 50-fold molar excess, calculated per dimer. The inclusion or absence of a filtering step is indicated by a “1” or a “2” symbol above each lane. The positions of the slot, protein–DNA complexes, and free DNA band are also indicated.
(TrpR): retention of protein–DNA complexes on nitrocellulose filters was not observed under standard conditions (39, 40) and can only be observed in the presence of high concentrations of ammonium sulfate (40). We therefore examined LacR 331Stop binding to plasmid DNA containing O id in our standard buffer containing in addition ammonium sulfate at various concentrations up to and including 2 M, a concentration that will fully precipitate LacR (32). However, binding was again not observed (data not show), indicating that the high salt conditions necessary for TrpR holorepressor (40) are not compatible with simultaneous binding of LacR 331Stop to nitrocellulose and DNA. The Residual Binding of DNA Observed with LacRBasic Is Due to the Basic Tail As noted above, weak binding is observed by LacRBasic at the repressor and operator concentrations used in Table 1. We therefore reexamined binding by titrating LacRBasic and LacR adi against plasmid DNA containing or not containing O id, in the presence or absence of 1 mM IPTG 4 in a filter-binding assay (Fig. 5). As expected, LacR adi shows a much higher affinity for plasmid DNA containing O id than for nonoperator DNA, and the observed operator-specific binding, but not operator-independent binding, is sensitive to IPTG. Binding by LacRBasic, however, is the same within experimental error to plasmid DNA regardless of whether the lac operator is present or not and regardless of whether inducer is added or not. This unspecific 4
Abbreviation used: IPTG, isopropyl b-D-thiogalactoside.
FIG. 5. The LacRBasic–DNA interaction detected by filter binding is unspecific. LacR adi or LacRBasic, at the indicated dimer concentrations, were mixed with plasmid pEE4 (nonoperator) or p310 (O idcontaining) DNA in the presence or absence of 1 mM IPTG and, after incubation, analyzed by filter binding. (a) LacR adi with pEE4 or p310. (b) LacRBasic with pEE4 or p310. Open squares indicate values obtained for binding to pEE4 in the absence of inducer. Filled diamonds indicate values obtained for binding to pEE4 in the presence of 1 mM IPTG. Open circles indicate values obtained for binding to p310 in the absence of inducer, and filled triangles indicate values obtained for binding to p310 in the presence of 1 mM IPTG. In the absence of protein, retention values of DNA on the filters were 2.6 6 1 and 1.1 6 0.3% for pEE4 and p310, respectively. Data points represent mean values 6 standard deviation for three independent determinations.
IS NITROCELLULOSE BINDING UNIVERSAL?
ular core, is required for nitrocellulose binding. This extends a previous observation that wild-type, tetrameric LacR no longer binds to nitrocellulose if it forms a loop complex with a DNA fragment containing two operators (27). It is now clear that binding of tetrameric LacR to a DNA fragment carrying a single operator can still be detected in filter-binding experiments because the second headpiece is still available for interactions with nitrocellulose. However, if a second operator is present (or if the DNA concentration is much higher than the protein concentration), then occupancy of the second headpiece by DNA simply removes it from availability for nitrocellulose binding. Consideration of the dimeric LacR variants we have examined here is also illuminating. First, LacRL 346A is only capable of forming dimers (29) because the single alanine substitution prevents formation of the fourhelical bundle that is responsible for tetramerization of the wild-type protein (41). This could lead either to the C-terminal extension simply becoming exposed to the solvent or to its becoming a random coil instead of forming an a-helix. Similarly, there is no reason to suppose that the relatively hydrophobic C-terminal extension introduced by a frame shift at position 330 in LacR adi (35) forms any definite structure. The lack of this C-terminal extension in LacR 331Stop is clearly responsible for its inability to bind simultaneously to nitrocellulose and DNA. In this context, it should be noted that the folding of the extreme C-terminus does not affect the folding of the LacR core, that is, residues prior to position 330 (29 –31, 35–38, 42). The lack of, or weak, binding with the remaining two variants we tested, LacR 331Gal and LacRBasic, gives us an additional insight into the nature of the extension required for nitrocellulose binding. In the case of LacR 331Gal, it is likely that the C-terminal 13 amino acids adopt their normal secondary structure when attached to the LacR core, thus neither being unfolded nor exposing a hydrophobic surface to the solvent. Even if this stretch of amino acids does not adopt its native structure in the LacR 331Gal context, note that it has an overall hydrophilic character. In the case of LacRBasic, additional experiments demonstrated that the weak binding observed in Table 1 is most probably due to unspecific DNA binding by the AKKKK repeat, as demonstrated previously (31), which leaves the headpiece free for nitrocellulose binding. Specific binding to O id by the headpiece of LacRBasic is not detectable by filter binding, either because the basic region immediately forms a loop or sandwich complex (cf. Ref. 31), thus removing the basic region from possible interactions with nitrocellulose, or because the structure of the basic region, almost certainly an a-helix (43) and obviously extremely hydrophilic by virtue of its highly positively charged nature, precludes it from binding to nitrocellulose. Although we cannot distinguish be-
335
tween these possibilities at present, the absence of evidence for loop formation by LacRBasic in vitro (31) makes the second possibility seem more likely. The LacR headpiece is not strictly speaking completely disordered in the non-DNA-bound form, although NMR structures of the isolated headpieces reveal that it does undergo a large structural transition between DNA-bound and nonbound forms (44, 45). Similarly, the headpieces are not visible in any X-ray structure determined in the absence of DNA (34) indicating a large structural flexibility at the very least. On the other hand, the core is a large, globular structure, the only other protrusion being the C-terminal 30 amino acids, which form a four-helical bundle (33, 34, 41). Our results indicate that it is the flexible nature of the headpieces, or alternatively the disordered nature of certain altered C-terminal extensions to the core, both most likely exposing hydrophobic regions to the solvent, that is responsible for LacR binding to nitrocellulose: the globular core itself has an entirely hydrophilic surface (33, 34) and is therefore incapable of binding to nitrocellulose. The Significance of the Mechanism of Nitrocellulose Binding A degree of structural disorder is exhibited by all DNA-binding proteins to date and may be of crucial importance for the way in which DNA-binding proteins locate and bind specific sites on DNA (46). The results presented here suggest this is also true for binding to nitrocellulose. Experience with TrpR suggests this property may not be limited to LacR. TrpR is a relatively small globular protein, of which the DNA-binding domain forms an integral part rather than being an extension as is the case for LacR (47, 48). DNA binding is only detected in filter-binding experiments in the presence of high concentrations of ammonium sulfate (40), most likely also because of the absence of a structurally disordered domain on the protein, in a manner analogous to LacR 331Stop. However, the underlying mechanism is not clear, given that there is also a DNA length dependence for TrpR complexes that may be detected, even under high salt conditions. These results sound a cautionary note indeed for prospective users of filter binding for detecting protein–DNA interactions. The use of filter binding is an integral part of the selection of specific binding sites from a library of random sequences for putative DNAbinding proteins whose target sequence is unknown, in a so-called binding site selection assay (49, 50). In this particular case, the immobilization of the protein on a column, or alternatively via magnetic beads, is preferable, in spite of the ease of handling offered by binding protein to a nitrocellulose disk. Similarly, for kinetic and thermodynamic analyses, it is recommended that
336
OEHLER, ALEX, AND BARKER
alternative approaches (19 –24) be used, particularly if the target DNAs used are short, unless it is clear that the protein in question is capable of binding to nitrocellulose and DNA simultaneously. The technique still has its advantages over other assays, but potential users should beware that a negative result obtained by filter binding does not necessarily mean that the protein in question cannot bind to the target DNA offered. ACKNOWLEDGMENTS We are grateful to Benno Mu¨ller-Hill for his encouragement, support, and numerous discussions throughout this study. We also appreciate the support of Martina Classen-Palm of Pall Gelman Sciences GmbH, including the gift of numerous differing membrane types. This study was supported by the GIF via grants to B.M.-H.
REFERENCES 1. 2. 3. 4. 5. 6. 7.
8. 9. 10. 11. 12. 13. 14. 15. 16.
17. 18. 19. 20. 21.
Ellford, W. J. (1931) J. Pathol. Bacteriol. 34, 505–535. Nirenberg, M., and Leder, P. (1964) Science 145, 1399 –1407. Jones, O. W., and Berg, P. (1966) J. Mol. Biol. 22, 199 –209. Riggs, A. D., Bourgeois, S., Newby, R. F., and Cohn, M. (1968) J. Mol. Biol. 34, 365–368. Riggs, A. D., Suzuki, H., and Bourgeois, S. (1970) J. Mol. Biol. 48, 67– 83. Towbin, H., Staehelin, T., and Gordon, J. (1979) Proc. Natl. Acad. Sci. USA 76, 4350 – 4354. Presswood, W. G. (1981) in Membrane Filtration: Applications, Techniques and Problems (Dutka, B. J., Ed.), pp. 1–17, Dekker, New York. Wallis, C., Melnick, J. L., and Gerba, C. P. (1979) Annu. Rev. Microbiol. 33, 413– 437. Gershoni, J. M., and Palade, G. E. (1983) Anal. Biochem. 131, 1–15. Woodbury, C. P., Jr., and von Hippel, P. H. (1983) Biochemistry 22, 4730 – 4737. Riggs, A. D., Bourgeois, S., and Cohn, M. (1970) J. Mol. Biol. 53, 401– 417. Berg, O. G., Winter, R. B., and von Hippel, P. H. (1981) Biochemistry 20, 6929 – 6948. Winter, R. B., and von Hippel, P. H. (1981) Biochemistry 20, 6948 – 6961. Winter, R. B., Berg, O. G., and von Hippel, P. H. (1981) Biochemistry 20, 6961– 6977. Record, M. T., Jr., Ha, J.-H., and Fisher, M. A. (1991) Methods Enzymol. 208, 291–343. Gilbert, W., Maxam, A., and Mirzabekov, A. (1976) in Control of Ribosome Synthesis, Alfred Benzon Symposium IX (Kjeldgaard, N. O., and Maaløe, O., Eds.), pp. 139 –148, Munksgaard, Copenhagen. Galas, D. J., and Schmitz, A. (1978) Nucleic Acids Res. 5, 3157– 3170. Siebenlist, U., and Gilbert, W. (1980) Proc. Natl. Acad. Sci. USA 77, 122–126. Sandaltzopoulos, R., and Becker, P. B. (1994) Nucleic Acids Res. 22, 1511–1512. Sandaltzopoulos, R., Quivy, J.-P., and Becker, P. B. (1995) Methods Mol. Cell. Biol. 5, 176 –181. Hsieh, M., and Brenowitz, M. (1996) Methods Enzymol. 274, 478 – 492.
22. Garner, M. M., and Revzin, A. (1981) Nucleic Acids Res. 9, 3047–3060. 23. Fried, M. G., and Crothers, D. M. (1981) Nucleic Acids Res. 9, 6505– 6525. 24. Vossen, K. M., and Fried, M. G. (1997) Anal. Biochem. 245, 85–92. 25. Heyduk, T., and Lee, J. C. (1990) Proc. Natl. Acad. Sci. USA 87, 1744 –1748. 26. Bondeson, K., Frostell-Karlsson, Å., Fa¨gerstam, L., and Magnusson, G. (1993) Anal. Biochem. 214, 245–251. 27. Fickert, R., and Mu¨ller-Hill, B. (1992) J. Mol. Biol. 226, 59 – 68. 28. Barker, A., Fickert, R., Oehler, S., and Mu¨ller-Hill, B. (1998) J. Mol. Biol. 278, 549 –558. 29. Alberti, S., Oehler, S., von Wilcken-Bergmann, B., Kra¨mer, H., and Mu¨ller-Hill, B. (1991) New Biologist 3, 57– 62. 30. Oehler, S., Amouyal, M., Kolkhof, P., von Wilcken-Bergmann, B., and Mu¨ller-Hill, B. (1994) EMBO J. 13, 3348 –3355. 31. Kolkhof, P., Oehler, S., Alex, R., and Mu¨ller-Hill, B. (1995) J. Mol. Biol. 247, 396 – 403. 32. Mu¨ller-Hill, B., Beyreuther, K., and Gilbert, W. (1971) Methods Enzymol. 21, 483– 487. 33. Friedman, A. M., Fischmann, T. O., and Steitz, T. A. (1995) Science 268, 1721–1727. 34. Lewis, M., Chang, G., Horton, N. C., Kercher, M. A., Pace, H. C., Schumacher, M. A., Brennan, R. G., and Lu, P. (1996) Science 271, 1247–1254. 35. Lehming, N., Sartorius, J., Oehler, S., von Wilcken-Bergmann, B., and Mu¨ller-Hill, B. (1988) Proc. Natl. Acad. Sci. USA 85, 7947–7951. 36. Oehler, S., Eismann, E. R., Kra¨mer, H., and Mu¨ller-Hill, B. (1990) EMBO J. 9, 973–979. 37. Chakerian, A. E., and Matthews, K. S. (1991) J. Biol. Chem. 266, 22206 –22214. 38. Brenowitz, M., Mandal, N., Pickar, A., Jamison, E., and Adhya, S. (1991) J. Biol. Chem. 266, 1281–1288. 39. Joachimiak, A., Kelley, R. L., Gunsalus, R. P., Yanofsky, C., and Sigler, P. (1983) Proc. Natl. Acad. Sci. USA 80, 668 – 672. 40. Klig, L. S., Crawford, I. P., and Yanofsky, C. (1987) Nucleic Acids Res. 15, 5339 –5351. 41. Li, L., and Matthews, K. S. (1995) J. Biol. Chem. 270, 10640 – 10649. 42. Alberti, S., Oehler, S., von Wilcken-Bergmann, B., and Mu¨llerHill, B. (1993) EMBO J. 12, 3227–3236. 43. Saha, S., Brickman, J. M., Lehming, N., and Ptashne, M. (1993) Nature 363, 648 – 652. 44. Slijper, M., Bonvin, A. M. J. J., Boelens, R., and Kaptein, R. (1996) J. Mol. Biol. 259, 761–773. 45. Slijper, M., Boelens, R., Davis, A. L., Konings, R. N. H., van der Marel, G. A., van Boom, J. H., and Kaptein, R. (1997) Biochemistry 36, 249 –254. 46. Barker, A., and Mu¨ller-Hill, B. (1998) FEBS Lett. 432, 1–3. 47. Schevitz, R. W., Otwinowski, Z., Joachimiak, A., Lawson, C. L., and Sigler, P. B. (1985) Nature 317, 782–786. 48. Gryk, M. R., Finucane, M. D., Zheng, Z., and Jardetzky, O. (1995) J. Mol. Biol. 246, 618 – 627. 49. Nørby, P. L., Pallisgaard, N., Pedersen, F. S., and Jørgensen, P. (1992) Nucleic Acids Res. 20, 6317– 6321. 50. Wright, W. E., and Funk, W. D. (1993) Trends Biochem. Sci. 18, 77– 80. 51. Beyreuther, K., Adler, K., Geisler, N., and Klemm, A. (1973) Proc. Natl. Acad. Sci. USA 70, 3576 –3580. 52. Farabaugh, P. J. (1978) Nature 274, 765–769.
Is Nitrocellulose Filter Binding Really a Universal Assay for Protein–DNA Interactions? Stefan Oehler, 1 Regina Alex, 2 and Andrew Barker 3 Institut fu¨r Genetik der Universita¨t zu Ko¨ln, Weyertal 121, 50931 Ko¨ln, Germany
Received September 24, 1998
The ability to bind to nitrocellulose is commonly accepted as being a universal property of proteins and has been widely used in many different fields of study. This property was first exploited in the study of DNAbinding proteins 30 years ago, in studies involving DNA binding by the lactose repressor (LacR) of Escherichia coli. Termed the filter-binding assay, it remains the quickest and easiest assay available for the study of protein–DNA interactions. However, the exact mechanism by which proteins bind to nitrocellulose remains uncertain. Given the supposedly universal nature of the interaction, we were surprised to notice that certain LacR variants were completely unable to bind simultaneously to DNA containing a single lac operator and nitrocellulose. Investigation of this loss of binding suggests that LacR requires a protein region that is both hydrophobic in nature and more or less unstructured, in order to bind to both nitrocellulose and DNA. In the case of wild-type, tetrameric LacR, the DNA-recognition domain that is not bound to DNA suffices. Dimeric LacR variants will only bind if they have certain C-terminal extensions. These experiments sound a cautionary note for the use of filter binding as an assay of choice, particularly in applications involving screening for the DNA-binding site of putative DNA-binding proteins. © 1999 Academic Press Key Words: filter-binding assay; nitrocellulose binding; mobility shift assay; protein–DNA recognition; Lac repressor.
The ability of proteins to bind to nitrocellulose has been recognized since at least the 1930s, when the 1 Present addresses: Department of Genetics, University of Cambridge, Downing St., Cambridge, CB2 3EH, England. 2 Present address: AMGEN GmbH, Reesstr. 25, 80992 Mu¨nchen, Germany. 3 To whom correspondence should be addressed. Fax: 149 221 470 5170. E-mail: [email protected].
330
necessity of using serum to prevent adsorption of virus particles to collodion (cellulose nitrate) membranes during filtration was described (1). In the 1960s, the discovery that ribosomes (2) and an RNA polymerase– DNA complex (3) bound to nitrocellulose filters rapidly led to the application of the property to other systems, including studies involving the DNA-binding of transcription factors (4, 5). The universal applicability of protein adsorption to nitrocellulose is perhaps best underlined by a relatively recent application, but one which is the most widely used: the electrophoretic transfer of denatured proteins to nitrocellulose for subsequent immunological characterization, commonly referred to as Western blotting (6). The exact mechanism, however, by which proteins bind to nitrocellulose remains unclear. Nitrocellulose is a fully nitrated derivative of cellulose, in which all free hydroxyl groups have been substituted by nitrate groups, and is thus hydrophobic in character. Nitrocellulose filters are structured rather like a sponge, through which individual pores follow tortuously winding paths and constitute about 80% of the total filter volume (7). It has long since been apparent that the process of protein binding involves active adsorption, as interacting particles such as viruses or ribosomes are clearly much smaller than the pores of the filter to which they were binding (2, 7, 8). Consideration of the conditions under which adsorption is optimal, and the chemical nature of nitrocellulose, led to the conclusion that binding requires hydrophobic interactions between protein and nitrocellulose, so it is not surprising that some proteins, notably those of small molecular weight, bind with reduced affinity (8 –10). As noted above, the nitrocellulose filter-binding assay has been widely used in analyzing both the thermodynamics and kinetics of protein–DNA interactions (5, 10 –15). Other assays have since become available. One of these is based on protection against base methylation or DNaseI cleavage by a bound protein, which 0003-2697/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved.
IS NITROCELLULOSE BINDING UNIVERSAL?
may be described under the generic name of footprinting (16 –18). The application of magnetic beads (19, 20) or of quenched-flow techniques (21) has enabled the handling times to be sufficiently shortened for the DNase protection assay to become amenable for kinetic studies. The near ubiquitous band shift assay (22, 23) has also proved applicable, at least with short target DNAs. The increased stability of protein–DNA complexes, once they have entered the gel matrix, means that this type of assay is also suitable for determination of kinetic dissociation rates (24). More recently, methods based on fluorescence anisotropy (25) or on plasmon resonance technology (26) have also appeared. These assays all have their strengths and weaknesses, a comparison of which is beyond the scope of this article. However, it may safely be said that the biggest advantages of filter binding in comparison to all other available methods are the ease and rapidity with which a large number of samples may be analyzed, expensive equipment is not required, and it is the assay of choice when analyzing protein binding to operator sequences contained in DNA molecules that are greater than a few kilobases in length. Previously, we reported the difference in association rate with operator located on a l DNA molecule between wild-type, tetrameric LacR, and a dimeric LacR variant, LacR adi (27, 28). We wished to examine this phenomenon further, with additional LacR variants that are also dimeric (29 –31). However, some of these proteins do not bind consecutively to DNA and nitrocellulose. We have undertaken a detailed examination of this loss of binding and present the results, and their implications for the study of protein–DNA interactions generally, here. MATERIALS AND METHODS
Protein and DNA Preparation LacR variants were purified as described previously (28, 32), with one exception. The LacRBasic variant exhibits a much tighter binding to phosphocellulose than other variants (data not shown) and therefore requires elution with a KPO 4 gradient of 0.075–1 M KPO 4, instead of a 0.075– 0.4 M KPO 4 gradient, as in the standard protocol (28). Otherwise, all steps in the purification were identical. Target DNAs were plasmids pEE4 (nonoperator) and p310 (containing O id), or, in the case of band shift assays, the 60- or 84-bp EcoRI–HindIII fragment of each plasmid, respectively. The plasmids, and methods used in purifying plasmid DNA, were as described previously (27, 28). For labeling, plasmid DNA was linearized with EcoRI, or with EcoRI–HindIII, and DNA labeled with [a- 32P]dATP (Amersham), by fill-in reaction using DNA Polymerase I Klenow fragment (Boehringer-Mannheim). Full-length plasmid DNA was then
331
separated from unincorporated label by passing over a Nick-column (Pharmacia), while the EcoRI–HindIII fragment was purified by electrophoresis on 6% nondenaturing acrylamide gels. Filter-Binding and Band Shift Assays In our standard assay (27), filter binding is performed with nitrocellulose disks 25 mm in diameter and with a mean pore size of 0.45 mm. Filters were pretreated by boiling for at least 5 min in distilled water and then, after washing in cold distilled water, allowed to equilibrate for at least 30 min in FB [10 mM Tris–HCl, pH 7.2, 0.1 mM EDTA, 3 mM Mg acetate, 10 mM KCl, 5% (v/v) dimethyl sulfoxide]. Binding reactions were performed in a final volume of 100 ml, in BB (FB containing in addition 0.1 mM dithiothreitol and 50mg/ml bovine serum albumin). Samples were allowed to equilibrate for 30 min before being filtered through a single filter disk (prewashed with 100 ml FB) with a filtering time of 5–10 s and washed with an additional 400 ml FB. Filters were allowed to air dry before being dissolved in 5 ml scintillant (Zinsser Analytik Quickszint 361) and counting. Reactions were performed with a DNA concentration of 5 3 10 211 M; protein concentration, calculated per dimer, was as indicated. Binding reactions for band shift assays were performed under identical conditions as for filter binding, except that the final reaction volume was 10 ml. Samples were mixed with 4 ml tracker dye [BB containing 0.04% (w/v) each bromophenol blue and xylene cyanol and 15% (w/v) Ficoll] and immediately loaded onto 6% polyacrylamide gels. Running buffer was 0.53 TBE (23) and gels were prerun at 20 V/cm until the initial resistance had decreased by at least half. Electrophoresis was performed for 2 h at 10 V/cm and at room temperature. After electrophoresis, gels were dried on Whatman 3M paper and exposed to X-ray film for detection of free and retarded DNA bands. Filtration of samples for analysis in band shift assays was performed as follows. Nitrocellulose disks of suitable size were pretreated as described above and then placed between a 5-ml syringe and the plastic base of a 19-gauge needle, from which the needle had first been removed. Samples in a total volume of 20 ml were pipetted directly on to the filter and then pushed through with the plunger at sufficient pressure to give a filtering time of about 1 s. In the case of protein–DNA mixtures, 10 ml of the filtrate was then processed for loading onto gels as described above. In the case of protein mixtures, samples were filtered at 23 final concentration and then used in binding reactions as described above. For reactions where samples were to be simultaneously analyzed by band shift and filter binding, the reaction was performed in a total volume of 110 ml. First, 10 ml was removed and loaded on
332
OEHLER, ALEX, AND BARKER
The results of a typical filter-binding experiment, examining the binding at equilibrium of these LacR variants to an 84-bp DNA fragment containing the ideal lac operator, O id, are shown in Table 1. A subset of dimeric LacR variants is no longer capable of recording a positive result in this test. Specifically, those variants in which the C-terminal extension is completely removed, or in which residues downstream of position 330 are replaced with the C-terminal 13 amino acids of GalR, appear to have lost the ability to bind to DNA, as determined by filter binding. Additionally, a LacR variant with an extremely basic extreme Cterminus only shows residual binding in this test (Table 1). The Apparent Loss of Binding Is a Loss of the Ability to Bind Simultaneously to Nitrocellulose and DNA
FIG. 1. Amino acid sequence of the extreme C-termini of the LacR variants used in this study: (a) wild-type, tetrameric LacR (51, 52); (b) LacRL 346A (29); (c) LacR adi (35); (d) LacR 331Stop (30); (e) LacR 331Gal (31); and (f) LacRBasic (31). The boxed residues in the sequence of wild-type LacR are those that comprise the leucine heptad repeat (29). Those residues that differ from wild-type LacR are underlined in the other sequences. Note that the substitution L 330S is neutral in wild-type LacR and, where it occurs, was introduced purely for convenience in DNA manipulation (30). The first residue shown in all cases is V 321.
polyacrylamide gels as described above, before the remainder was filtered through nitrocellulose. RESULTS
Apparent Loss of DNA Binding in Filter-Binding Experiments LacR consists of three structural domains: an Nterminal extension (amino acids 1– 61) that comprises the DNA-binding domain; a globular core (amino acids 62–330) that comprises the bulk of the protein, which contains the inducer-binding domain and the dimerization domain between monomers; and a C-terminal domain (amino acids 331–360) that includes two leucine heptad repeats which allow two dimers to aggregate to a tetramer (29, 33, 34). Mutations which disrupt the heptad repeat lead to dimeric forms of LacR, which retain DNA-binding activity but have a dramatically reduced ability to bind cooperatively to two operators (29 –31, 35–38). The amino acid sequences of the extreme C-termini of wild-type, tetrameric LacR and a number of dimeric LacR variants are shown in Fig. 1. As all variants are identical in both DNA-binding and core domains, they should be unaltered in binding affinity, at least to a short DNA fragment carrying a single lac operator.
The same binding mixes assayed by filter binding in Table 1 were also examined in band shift assays (Fig. 2). It is immediately apparent that all LacR variants examined are capable of binding to the DNA fragment in this experiment. Thus, the result presented in Table 1 does not reflect an inability of the LacR variants in question to bind DNA, but an inability to bind simultaneously to DNA and to nitrocellulose. This can also be demonstrated by passing the binding reactions through a nitrocellulose filter before loading on an acrylamide gel. The retarded band (but not the free DNA band) is removed if the binding reaction is filtered before loading for wild-type LacR, LacRL 346A, and LacR adi, but not for LacR 331Stop, LacR 331Gal, or LacRBasic (Fig. 2). A detailed examination of this phenomenon is presented in Fig. 3, where two dimeric LacR variants, LacR adi and LacR 331Stop, are shown as representative nitrocellulose-binding and nonbinding variants, respectively. Both variants were titrated against a fixed
TABLE 1
DNA-Binding Activity of LacR Variants as Determined in Filter-Binding Assays LacR variant a
% input DNA retained on nitrocellulose filters b
None Wild-type LacR LacRL 346A LacR adi LacR 331Stop LacR 331Gal LacRBasic
0.5 6 0.3 24 6 0.9 48 6 2.6 48 6 4.4 0.7 6 0.3 1.2 6 0.6 7.8 6 0.4
a Purified LacR was added at a 50-fold molar excess, calculated per LacR dimer. Target DNA is identical to that used in Fig. 2. b Retention of 5% or less of the input target DNA on nitrocellulose filters is scored as nonbinding (5, 11, 27).
IS NITROCELLULOSE BINDING UNIVERSAL?
FIG. 2. All LacR variants bind DNA but not all bind DNA and nitrocellulose consecutively. Binding reactions, with a 50-fold molar excess of protein, calculated per dimer, were analyzed in a band shift assay. The LacR variants examined are indicated above the figure. In each pair of reactions, the lane on the left was loaded directly after incubation, while the lane on the right was subjected to filtration through nitrocellulose immediately before loading. This is indicated by the presence of a “2” or a “1,” respectively, above each lane and the word “filtered” to the left of the figure. The positions of the slot, protein–DNA complexes, and the free DNA band are also indicated.
333
We undertook additional control experiments to ensure that the lack of binding we observe is not due to some special property of the particular type (Sartorius, mean pore size 0.45 mm) of nitrocellulose filter used routinely in this laboratory. Again, LacR 331Stop and LacR adi were used as representative nonbinding and binding LacR variants, respectively. Identical results to those presented above were obtained with nitrocellulose of a comparable mean pore size manufactured by Millipore, Amersham, Schleicher & Schuell, and Gelman Sciences and Sartorius nitrocellulose filters with a mean pore size of 0.22 mm or other membrane types also available for protein immobilization (data not shown). The results presented here are reminiscent of those obtained with the Escherichia coli Trp repressor
concentration of target DNA. In band shift assays, binding is observed when an excess of either LacR adi or LacR 331Stop is added (Fig. 3a), but again, saturation binding of the operator fragment is only observed for LacR adi if the samples are analyzed by filter binding (Fig. 3b). Thus, the lack of nitrocellulose binding observed for LacR 331Stop is independent of the amount of protein added. LacR Needs a Disordered Domain for Nitrocellulose Binding A simple explanation for the inability of certain LacR variants to bind simultaneously to nitrocellulose and DNA is that both are bound by the same protein domain, namely the DNA-binding headpiece. To test this idea, we performed a band shift assay using the same LacR variants shown in Fig. 2, in which the protein was either untreated or filtered through nitrocellulose filters prior to addition to DNA. The result, shown in Fig. 4, clearly demonstrates that all variants tested bind to nitrocellulose. After preincubation with DNA, some dimeric LacR variants fail to bind to nitrocellulose (Table 1, Fig. 2, Fig. 3). Thus, in these cases, the only protein domain capable of nitrocellulose binding is the headpiece. Those LacR variants still capable of binding simultaneously to nitrocellulose and operator DNA are those in which there is more than one such domain—the second, non-DNA-bound headpiece in the case of wild-type, tetrameric LacR or the (presumably) disordered C-terminal extension in the case of LacRL 346A and LacR adi.
FIG. 3. The lack of nitrocellulose binding by LacR 331Stop is not dependent on protein concentration. (a) Band shift assay performed with LacR adi and LacR 331Stop. The LacR variants used are indicated above the gel, and increasing protein concentration is indicated by a wedge. Lanes in which no protein was added are indicated by a “(2)” sign. The positions of protein–DNA complexes and free DNA are indicated to the left of the gel. (b) Filter-binding assay performed with the same samples. Data points for LacR adi are indicated by open circles, while data points for LacR 331Stop are indicated by filled squares. Protein concentration, calculated per dimer, is indicated. In the absence of protein, DNA retention on the filters was 5 6 0.5% of the input DNA in this experiment.
334
OEHLER, ALEX, AND BARKER
binding must be due to the basic tail of LacRBasic, with the non-DNA-bound headpieces left available for nitrocellulose binding. DISCUSSION
The Significance of the LacR–Nitrocellulose Interaction We have shown that the reason why certain LacR variants cannot bind simultaneously to DNA and nitrocellulose is that the presence of at least one freely accessible structural domain, protruding from the glob-
FIG. 4. LacR alone is capable of binding to nitrocellulose. A band shift assay is shown in which the indicated LacR variants were either mixed directly with target DNA or passed through nitrocellulose filters immediately prior to mixing with target DNA. Protein was added at a 50-fold molar excess, calculated per dimer. The inclusion or absence of a filtering step is indicated by a “1” or a “2” symbol above each lane. The positions of the slot, protein–DNA complexes, and free DNA band are also indicated.
(TrpR): retention of protein–DNA complexes on nitrocellulose filters was not observed under standard conditions (39, 40) and can only be observed in the presence of high concentrations of ammonium sulfate (40). We therefore examined LacR 331Stop binding to plasmid DNA containing O id in our standard buffer containing in addition ammonium sulfate at various concentrations up to and including 2 M, a concentration that will fully precipitate LacR (32). However, binding was again not observed (data not show), indicating that the high salt conditions necessary for TrpR holorepressor (40) are not compatible with simultaneous binding of LacR 331Stop to nitrocellulose and DNA. The Residual Binding of DNA Observed with LacRBasic Is Due to the Basic Tail As noted above, weak binding is observed by LacRBasic at the repressor and operator concentrations used in Table 1. We therefore reexamined binding by titrating LacRBasic and LacR adi against plasmid DNA containing or not containing O id, in the presence or absence of 1 mM IPTG 4 in a filter-binding assay (Fig. 5). As expected, LacR adi shows a much higher affinity for plasmid DNA containing O id than for nonoperator DNA, and the observed operator-specific binding, but not operator-independent binding, is sensitive to IPTG. Binding by LacRBasic, however, is the same within experimental error to plasmid DNA regardless of whether the lac operator is present or not and regardless of whether inducer is added or not. This unspecific 4
Abbreviation used: IPTG, isopropyl b-D-thiogalactoside.
FIG. 5. The LacRBasic–DNA interaction detected by filter binding is unspecific. LacR adi or LacRBasic, at the indicated dimer concentrations, were mixed with plasmid pEE4 (nonoperator) or p310 (O idcontaining) DNA in the presence or absence of 1 mM IPTG and, after incubation, analyzed by filter binding. (a) LacR adi with pEE4 or p310. (b) LacRBasic with pEE4 or p310. Open squares indicate values obtained for binding to pEE4 in the absence of inducer. Filled diamonds indicate values obtained for binding to pEE4 in the presence of 1 mM IPTG. Open circles indicate values obtained for binding to p310 in the absence of inducer, and filled triangles indicate values obtained for binding to p310 in the presence of 1 mM IPTG. In the absence of protein, retention values of DNA on the filters were 2.6 6 1 and 1.1 6 0.3% for pEE4 and p310, respectively. Data points represent mean values 6 standard deviation for three independent determinations.
IS NITROCELLULOSE BINDING UNIVERSAL?
ular core, is required for nitrocellulose binding. This extends a previous observation that wild-type, tetrameric LacR no longer binds to nitrocellulose if it forms a loop complex with a DNA fragment containing two operators (27). It is now clear that binding of tetrameric LacR to a DNA fragment carrying a single operator can still be detected in filter-binding experiments because the second headpiece is still available for interactions with nitrocellulose. However, if a second operator is present (or if the DNA concentration is much higher than the protein concentration), then occupancy of the second headpiece by DNA simply removes it from availability for nitrocellulose binding. Consideration of the dimeric LacR variants we have examined here is also illuminating. First, LacRL 346A is only capable of forming dimers (29) because the single alanine substitution prevents formation of the fourhelical bundle that is responsible for tetramerization of the wild-type protein (41). This could lead either to the C-terminal extension simply becoming exposed to the solvent or to its becoming a random coil instead of forming an a-helix. Similarly, there is no reason to suppose that the relatively hydrophobic C-terminal extension introduced by a frame shift at position 330 in LacR adi (35) forms any definite structure. The lack of this C-terminal extension in LacR 331Stop is clearly responsible for its inability to bind simultaneously to nitrocellulose and DNA. In this context, it should be noted that the folding of the extreme C-terminus does not affect the folding of the LacR core, that is, residues prior to position 330 (29 –31, 35–38, 42). The lack of, or weak, binding with the remaining two variants we tested, LacR 331Gal and LacRBasic, gives us an additional insight into the nature of the extension required for nitrocellulose binding. In the case of LacR 331Gal, it is likely that the C-terminal 13 amino acids adopt their normal secondary structure when attached to the LacR core, thus neither being unfolded nor exposing a hydrophobic surface to the solvent. Even if this stretch of amino acids does not adopt its native structure in the LacR 331Gal context, note that it has an overall hydrophilic character. In the case of LacRBasic, additional experiments demonstrated that the weak binding observed in Table 1 is most probably due to unspecific DNA binding by the AKKKK repeat, as demonstrated previously (31), which leaves the headpiece free for nitrocellulose binding. Specific binding to O id by the headpiece of LacRBasic is not detectable by filter binding, either because the basic region immediately forms a loop or sandwich complex (cf. Ref. 31), thus removing the basic region from possible interactions with nitrocellulose, or because the structure of the basic region, almost certainly an a-helix (43) and obviously extremely hydrophilic by virtue of its highly positively charged nature, precludes it from binding to nitrocellulose. Although we cannot distinguish be-
335
tween these possibilities at present, the absence of evidence for loop formation by LacRBasic in vitro (31) makes the second possibility seem more likely. The LacR headpiece is not strictly speaking completely disordered in the non-DNA-bound form, although NMR structures of the isolated headpieces reveal that it does undergo a large structural transition between DNA-bound and nonbound forms (44, 45). Similarly, the headpieces are not visible in any X-ray structure determined in the absence of DNA (34) indicating a large structural flexibility at the very least. On the other hand, the core is a large, globular structure, the only other protrusion being the C-terminal 30 amino acids, which form a four-helical bundle (33, 34, 41). Our results indicate that it is the flexible nature of the headpieces, or alternatively the disordered nature of certain altered C-terminal extensions to the core, both most likely exposing hydrophobic regions to the solvent, that is responsible for LacR binding to nitrocellulose: the globular core itself has an entirely hydrophilic surface (33, 34) and is therefore incapable of binding to nitrocellulose. The Significance of the Mechanism of Nitrocellulose Binding A degree of structural disorder is exhibited by all DNA-binding proteins to date and may be of crucial importance for the way in which DNA-binding proteins locate and bind specific sites on DNA (46). The results presented here suggest this is also true for binding to nitrocellulose. Experience with TrpR suggests this property may not be limited to LacR. TrpR is a relatively small globular protein, of which the DNA-binding domain forms an integral part rather than being an extension as is the case for LacR (47, 48). DNA binding is only detected in filter-binding experiments in the presence of high concentrations of ammonium sulfate (40), most likely also because of the absence of a structurally disordered domain on the protein, in a manner analogous to LacR 331Stop. However, the underlying mechanism is not clear, given that there is also a DNA length dependence for TrpR complexes that may be detected, even under high salt conditions. These results sound a cautionary note indeed for prospective users of filter binding for detecting protein–DNA interactions. The use of filter binding is an integral part of the selection of specific binding sites from a library of random sequences for putative DNAbinding proteins whose target sequence is unknown, in a so-called binding site selection assay (49, 50). In this particular case, the immobilization of the protein on a column, or alternatively via magnetic beads, is preferable, in spite of the ease of handling offered by binding protein to a nitrocellulose disk. Similarly, for kinetic and thermodynamic analyses, it is recommended that
336
OEHLER, ALEX, AND BARKER
alternative approaches (19 –24) be used, particularly if the target DNAs used are short, unless it is clear that the protein in question is capable of binding to nitrocellulose and DNA simultaneously. The technique still has its advantages over other assays, but potential users should beware that a negative result obtained by filter binding does not necessarily mean that the protein in question cannot bind to the target DNA offered. ACKNOWLEDGMENTS We are grateful to Benno Mu¨ller-Hill for his encouragement, support, and numerous discussions throughout this study. We also appreciate the support of Martina Classen-Palm of Pall Gelman Sciences GmbH, including the gift of numerous differing membrane types. This study was supported by the GIF via grants to B.M.-H.
REFERENCES 1. 2. 3. 4. 5. 6. 7.
8. 9. 10. 11. 12. 13. 14. 15. 16.
17. 18. 19. 20. 21.
Ellford, W. J. (1931) J. Pathol. Bacteriol. 34, 505–535. Nirenberg, M., and Leder, P. (1964) Science 145, 1399 –1407. Jones, O. W., and Berg, P. (1966) J. Mol. Biol. 22, 199 –209. Riggs, A. D., Bourgeois, S., Newby, R. F., and Cohn, M. (1968) J. Mol. Biol. 34, 365–368. Riggs, A. D., Suzuki, H., and Bourgeois, S. (1970) J. Mol. Biol. 48, 67– 83. Towbin, H., Staehelin, T., and Gordon, J. (1979) Proc. Natl. Acad. Sci. USA 76, 4350 – 4354. Presswood, W. G. (1981) in Membrane Filtration: Applications, Techniques and Problems (Dutka, B. J., Ed.), pp. 1–17, Dekker, New York. Wallis, C., Melnick, J. L., and Gerba, C. P. (1979) Annu. Rev. Microbiol. 33, 413– 437. Gershoni, J. M., and Palade, G. E. (1983) Anal. Biochem. 131, 1–15. Woodbury, C. P., Jr., and von Hippel, P. H. (1983) Biochemistry 22, 4730 – 4737. Riggs, A. D., Bourgeois, S., and Cohn, M. (1970) J. Mol. Biol. 53, 401– 417. Berg, O. G., Winter, R. B., and von Hippel, P. H. (1981) Biochemistry 20, 6929 – 6948. Winter, R. B., and von Hippel, P. H. (1981) Biochemistry 20, 6948 – 6961. Winter, R. B., Berg, O. G., and von Hippel, P. H. (1981) Biochemistry 20, 6961– 6977. Record, M. T., Jr., Ha, J.-H., and Fisher, M. A. (1991) Methods Enzymol. 208, 291–343. Gilbert, W., Maxam, A., and Mirzabekov, A. (1976) in Control of Ribosome Synthesis, Alfred Benzon Symposium IX (Kjeldgaard, N. O., and Maaløe, O., Eds.), pp. 139 –148, Munksgaard, Copenhagen. Galas, D. J., and Schmitz, A. (1978) Nucleic Acids Res. 5, 3157– 3170. Siebenlist, U., and Gilbert, W. (1980) Proc. Natl. Acad. Sci. USA 77, 122–126. Sandaltzopoulos, R., and Becker, P. B. (1994) Nucleic Acids Res. 22, 1511–1512. Sandaltzopoulos, R., Quivy, J.-P., and Becker, P. B. (1995) Methods Mol. Cell. Biol. 5, 176 –181. Hsieh, M., and Brenowitz, M. (1996) Methods Enzymol. 274, 478 – 492.
22. Garner, M. M., and Revzin, A. (1981) Nucleic Acids Res. 9, 3047–3060. 23. Fried, M. G., and Crothers, D. M. (1981) Nucleic Acids Res. 9, 6505– 6525. 24. Vossen, K. M., and Fried, M. G. (1997) Anal. Biochem. 245, 85–92. 25. Heyduk, T., and Lee, J. C. (1990) Proc. Natl. Acad. Sci. USA 87, 1744 –1748. 26. Bondeson, K., Frostell-Karlsson, Å., Fa¨gerstam, L., and Magnusson, G. (1993) Anal. Biochem. 214, 245–251. 27. Fickert, R., and Mu¨ller-Hill, B. (1992) J. Mol. Biol. 226, 59 – 68. 28. Barker, A., Fickert, R., Oehler, S., and Mu¨ller-Hill, B. (1998) J. Mol. Biol. 278, 549 –558. 29. Alberti, S., Oehler, S., von Wilcken-Bergmann, B., Kra¨mer, H., and Mu¨ller-Hill, B. (1991) New Biologist 3, 57– 62. 30. Oehler, S., Amouyal, M., Kolkhof, P., von Wilcken-Bergmann, B., and Mu¨ller-Hill, B. (1994) EMBO J. 13, 3348 –3355. 31. Kolkhof, P., Oehler, S., Alex, R., and Mu¨ller-Hill, B. (1995) J. Mol. Biol. 247, 396 – 403. 32. Mu¨ller-Hill, B., Beyreuther, K., and Gilbert, W. (1971) Methods Enzymol. 21, 483– 487. 33. Friedman, A. M., Fischmann, T. O., and Steitz, T. A. (1995) Science 268, 1721–1727. 34. Lewis, M., Chang, G., Horton, N. C., Kercher, M. A., Pace, H. C., Schumacher, M. A., Brennan, R. G., and Lu, P. (1996) Science 271, 1247–1254. 35. Lehming, N., Sartorius, J., Oehler, S., von Wilcken-Bergmann, B., and Mu¨ller-Hill, B. (1988) Proc. Natl. Acad. Sci. USA 85, 7947–7951. 36. Oehler, S., Eismann, E. R., Kra¨mer, H., and Mu¨ller-Hill, B. (1990) EMBO J. 9, 973–979. 37. Chakerian, A. E., and Matthews, K. S. (1991) J. Biol. Chem. 266, 22206 –22214. 38. Brenowitz, M., Mandal, N., Pickar, A., Jamison, E., and Adhya, S. (1991) J. Biol. Chem. 266, 1281–1288. 39. Joachimiak, A., Kelley, R. L., Gunsalus, R. P., Yanofsky, C., and Sigler, P. (1983) Proc. Natl. Acad. Sci. USA 80, 668 – 672. 40. Klig, L. S., Crawford, I. P., and Yanofsky, C. (1987) Nucleic Acids Res. 15, 5339 –5351. 41. Li, L., and Matthews, K. S. (1995) J. Biol. Chem. 270, 10640 – 10649. 42. Alberti, S., Oehler, S., von Wilcken-Bergmann, B., and Mu¨llerHill, B. (1993) EMBO J. 12, 3227–3236. 43. Saha, S., Brickman, J. M., Lehming, N., and Ptashne, M. (1993) Nature 363, 648 – 652. 44. Slijper, M., Bonvin, A. M. J. J., Boelens, R., and Kaptein, R. (1996) J. Mol. Biol. 259, 761–773. 45. Slijper, M., Boelens, R., Davis, A. L., Konings, R. N. H., van der Marel, G. A., van Boom, J. H., and Kaptein, R. (1997) Biochemistry 36, 249 –254. 46. Barker, A., and Mu¨ller-Hill, B. (1998) FEBS Lett. 432, 1–3. 47. Schevitz, R. W., Otwinowski, Z., Joachimiak, A., Lawson, C. L., and Sigler, P. B. (1985) Nature 317, 782–786. 48. Gryk, M. R., Finucane, M. D., Zheng, Z., and Jardetzky, O. (1995) J. Mol. Biol. 246, 618 – 627. 49. Nørby, P. L., Pallisgaard, N., Pedersen, F. S., and Jørgensen, P. (1992) Nucleic Acids Res. 20, 6317– 6321. 50. Wright, W. E., and Funk, W. D. (1993) Trends Biochem. Sci. 18, 77– 80. 51. Beyreuther, K., Adler, K., Geisler, N., and Klemm, A. (1973) Proc. Natl. Acad. Sci. USA 70, 3576 –3580. 52. Farabaugh, P. J. (1978) Nature 274, 765–769.