Experimental Strategies For The Identification Of Dna Binding Proteins

Published Date: 02 Nov 2017

Several methods have been used to detect and characterize protein–DNA interactions in vitro, including EMSA, DNase I footprinting, exonuclease III footprinting, southwestern blotting, various chemical protection and interference assays, and ultraviolet cross-linking. However, EMSA is by far the most frequently used, largely because it is technically the easiest and is often the most sensitive.

Briefly, the EMSA is based on the principle that a protein–DNA complex migrates through a native gel more slowly than free DNA, with the mobility of the complex determined by the size, shape, charge, and multimeric state of the protein. Thus, proteins within a crude cell extract that specifically recognize a given control element can be identified by incubating a small radiolabeled DNA fragment with the extract to allow the formation of protein–DNA complexes. The mixture is then electrophoresed through a native polyacrylamide gel, which separates the free radiolabeled probe DNA from the molecules bound by proteins. The free and bound DNA molecules are detected by autoradiography or Phosphorimager analysis.

Figure 1.

First and foremost, the EMSA is more sensitive and can reveal a specific protein–DNA complex even when the protein is at a low concentration within the extract. This is because a protein bound to only a few percent of the probe molecules still results in a complex that migrates more slowly than the free probe. If the specific activity of the probe is sufficiently high, this complex can easily be detected by autoradiography or Phosphorimager analysis.

-The EMSA requires the protein–DNA interaction to be maintained during gel electrophoresis.

-The amount of nonspecific competitor DNA typically needed for the assay when analyzing a crude cell extract. In both assays, an excess of a competitor DNA such as poly(dI:dC) or poly(dA:dT) is usually included to prevent nonspecific and low-specificity DNA-binding proteins from binding the probe and obscuring the specific complexes.

-Relatively high concentrations of competitor are needed for EMSA reactions because modest amounts of nonspecific nucleic-acid-binding proteins can lead to a smear of radioactive probe on the gel image, which prevents detection of the desired complexes.

-The high concentrations of competitor used for EMSA reactions may compete with the probe for binding of the protein of interest, diminishing or abolishing the specific protein–DNA complex.

-Although most protein–DNA interactions are unaffected or even enhanced by the resulting ions, some interactions may be weakened.

-For some DNA-binding proteins, short oligonucleotides of 20–25 bp do not support stable protein–DNA interactions, making it necessary to use longer probes. The failure of short probes could result from the protein's requirement for a longer stretch of adjacent DNA for stable binding. Alternatively, short probes containing a particular recognition sequence might anneal improperly, dissociate during the binding reaction, or possess secondary structures that prevent binding. Several methods can be used to generate longer probes, one of which is simply to use longer synthetic oligonucleotides. Another method is to excise and radiolabel a restriction fragment from a plasmid containing the DNA element.

-If attempts to detect DNA-binding activities by EMSA fail, the nuclear extract can be further concentrated or fractionated to enhance the probability of success. The protein concentration of the extract can be increased by ammonium sulfate precipitation followed by extensive dialysis. A DNA-binding activity of interest may be detected more readily using a more concentrated extract. Dignam and Roeder found that extract precipitated using 53% ammonium sulfate was optimal to enhance the activity of in vitro transcription extracts, but a different percentage may be optimal for a particular DNA-binding protein.

A second strategy to improve the detection of DNA-binding activities is to fractionate the extract (Ausubel et al. 1994, Unit 10.10). The negatively charged affinity resins heparin agarose and heparin sepharose are useful for this purpose. If extracts are applied to columns containing these resins in a buffer containing 0.1 m KCl or NaCl, most DNA-binding proteins will elute with a 0.4 m salt step. A 1 m salt step elution should also be performed in case the protein binds the resin with an unusually high affinity. The fractionation procedure separates the DNA-binding proteins from other proteins that might inhibit a binding reaction. Furthermore, the specific proteins within the 0.4 and 1 m salt fractions may be more concentrated than in the crude extracts. Therefore, DNA-binding activities may be detected in these fractions that were not detected in the crude extract.

-ompetitor DNA and Other Components of Binding Reactions

The nonspecific competitor DNA added to the binding reaction can be critical for the success of an EMSA experiment, especially when a crude nuclear extract is used. The competitor DNA prevents nonspecific and low-specificity nucleic acid-binding proteins from interacting stably with the radiolabeled probe. In the absence of an appropriate nonspecific competitor, numerous proteins within the extract (including abundant RNA-binding proteins) will bind the probe, leading to a radioactive smear on the gel image. Increased concentrations of competitor sequester more protein molecules from the radiolabeled probe, leaving primarily those proteins that have the highest specific affinity for the probe relative to the unlabeled competitor. Although high competitor concentrations often enhance the quality and clarity of the results, they also increase the probability that the relevant DNA-binding protein will be sequestered from the labeled probe.

The most frequently used nonspecific competitor is poly(dI:dC). However, other competitors, including poly(dG:dC), poly(dA:dT), or sheared genomic DNA from E. coli or salmon sperm, may yield better (or at least different) results. If the protein of interest binds the competitor with high affinity, it might not yield a detectable complex on the EMSA gel. Instead, complexes containing proteins that are not relevant to the function of the control element may be observed if those proteins bind the competitor with a lower affinity. Therefore, it can be beneficial to test different competitors when attempting to identify candidates for the relevant regulator of a DNA element.

In addition to the competitor DNA, other components of the binding reactions can be varied to enhance detection of protein–DNA complexes, such as the concentration of monovalent or divalent cations. Some protein–DNA complexes may benefit from inclusion of a low concentration (e.g., 0.01%) of the nonionic detergent Nonidet P-40 in the binding reaction. Others may benefit from inclusion of polyvinyl alcohol (2%), which can increase the effective concentration of proteins in the extract. For additional recommendations for optimizing binding reactions for EMSA experiments, see Ausubel et al. (1994, Unit 12.2).

Gel Electrophoresis Conditions

When developing an EMSA for a DNA element of interest, pilot experiments with different gel electrophoresis buffers are strongly recommended, as detection of some protein–DNA complexes depends on the specific buffer used. The three most common EMSA gel buffers are Tris-borate-EDTA, Tris-acetate-EDTA, and Tris-glycine. Inclusion of glycerol (5%) can also enhance the detection of some binding activities. Furthermore, electrophoresis can be performed at room temperature or at 4°C. For more information about different gel electrophoresis buffers, see Ausubel et al. (1994, Unit 12.2).

Design of Initial Experiments and Interpretation of Results

Initial EMSA experiments performed with a radiolabeled probe containing a control element of interest can result in the detection of one or more protein–DNA complexes on the gel image. Initial experiments are usually performed with a few different extract amounts (between 2 and 20 µg) and, for each extract amount, different amounts of nonspecific competitor (1–10 µg). If complexes are not detected, the binding reaction and gel conditions can be modified as noted above. For each complex detected, two questions are initially addressed: (1) Is the protein–DNA complex specific for the probe being tested? (2) Is there a significant probability that the complex is responsible for the in vivo function of the DNA element in the context of the endogenous control region?

These questions are generally addressed by additional EMSA experiments performed in the presence of specific competitor DNAs or with mutant probes. The most straightforward information is usually provided by comparing the complexes obtained using a radiolabeled wild-type probe with those obtained using a radiolabeled mutant probe. If a protein–DNA complex is not observed with a probe containing nucleotide substitutions in the control element of interest, the protein–DNA interaction almost certainly involves those nucleotides. This answers both questions posed above, at least in part, by demonstrating that the interaction is sequence-specific and involves functionally relevant nucleotides. Further evidence that the complex might be responsible for the function of the control element can be obtained by preparing probes that contain other substitution mutations. If nucleotide substitutions that selectively disrupt the function of the control element (e.g., in a transfection experiment) correspond to those that disrupt the EMSA complex, the probability that the complex is functionally relevant is enhanced.

As with all standard EMSA experiments, the mutant analysis should be performed with low concentrations of probe (1–10 fmoles per reaction) and protein concentrations that yield complexes within a linear range (i.e., 1–50% of the probe associated with protein). Comparisons using wild-type and mutant probes should be performed with radiolabeled probes possessing similar specific activities. If the mutant probe has a lower specific activity than the wild-type probe, the protein–DNA complex will appear to be less abundant, even if the mutation has no effect on the affinity of the interaction. Unfortunately, the relative affinities of a protein for wild-type and mutant sequences cannot be determined simply by comparing the abundance of a complex formed with wild-type and mutant probes; slight decreases in affinity can lead to dramatic decreases in the amount of complex detected if the complex is not sufficiently stable to survive gel electrophoresis. Thus, if a mutation diminishes the abundance of a complex, one can conclude only that the protein has a lower affinity for the mutant probe than for the wild-type probe, but the magnitude of the difference cannot be discerned. On the other hand, if a mutation has no effect on complex formation, the protein most likely binds the wild-type and mutant probes with comparable affinities. Overall, despite the lack of quantitative data, a comparison of wild-type and mutant probes can provide valuable information about (1) the specificity of an interaction, (2) the nucleotides required for the interaction, and (3) the relationship between the required nucleotides and those required for function of the DNA element.

Another approach to examine the specificity and functional relevance of a protein–DNA complex is to perform competition experiments with unlabeled wild-type and mutant DNA fragments, such as annealed oligonucleotides, PCR products, or restriction fragments. For the results of competition experiments to be meaningful, the wild-type and mutant competitors must be quantified carefully to ensure that their concentrations are comparable. Competition experiments are performed by including a radiolabeled wild-type probe in a series of binding reactions containing increasing concentrations of unlabeled wild-type or mutant competitor DNAs. Radiolabeled complex abundance decreases in the presence of unlabeled competitor because the DNA-binding protein distributes itself randomly among the probe and competitor molecules. Because the wild-type competitor has a higher affinity for the protein than does the mutant competitor, lower concentrations are required to reduce the abundance of the complex.

Valuable information about the sequence specificity and functional relevance of a protein–DNA interaction can be obtained by comparing a wild-type competitor to a series of mutant competitors. If an EMSA complex is diminished more by the wild-type competitor than by a mutant, the protein must be binding DNA in a sequence-specific manner. If the mutations that diminish competition correspond to the mutations that reduce the function of the DNA element (e.g., in a transfection assay), the results suggest that the protein–DNA interaction may be functionally relevant.

The principal advantage of assessing specificity by using mutant competitor DNAs instead of radiolabeled mutant probes is that the former provides more accurate information about the relative affinities of a protein for the wild-type and mutant sequences. As explained above, a comparison of radiolabeled wild-type and mutant probes provides limited information about relative binding affinities. Competition experiments are not subject to the same caveat because the key events involved in the comparison of wild-type and mutant sequences occur in solution during the initial binding reaction; gel electrophoresis is performed after the competition has been completed. If the affinity of the protein for a mutant sequence is twofold less than its affinity for a wild-type sequence, then a twofold higher concentration of the mutant competitor relative to the wild-type competitor should reduce complex abundance by a given amount. Thus, the competition strategy has greater potential to provide meaningful quantitative information about relative affinities.

Although the above considerations appear to be straightforward, the results of competition experiments can be complicated. For example, a twofold reduction often requires a large excess of wild-type competitor. This is because most EMSA experiments contain excess protein, despite the presence of free probe at the bottom of the gel. Thus, competition occurs only when the concentration of competitor is high enough to exceed the Kd of the interaction. Although this issue makes it difficult to predict the effect of a given concentration of wild-type competitor, it does not completely invalidate the comparison of wild-type and mutant competitors. Even if a 50-fold excess of the wild-type competitor is needed to achieve a twofold reduction in the abundance of a protein–DNA complex, the effect of the mutant competitor can provide insight into the relative affinities of the protein for the wild-type and mutant sequences. If a 50-fold excess of the mutant competitor results in a similar twofold reduction in the protein–DNA complex, one can conclude that the mutation has no significant effect on the affinity of the protein for the DNA. If a 100-fold excess of the mutant competitor is needed for a twofold reduction in the protein–DNA complex, the affinity of the protein for the mutant sequence is approximately twofold less than that of the wild-type sequence. Because the radiolabeled probe and competition strategies for assessing specificity and functional relevance of a protein–DNA interaction possess unique advantages and limitations, the use of both is strongly recommended.

Analysis of EMSA Complexes by Methylation Interference

The experiments discussed above provide information about the specificity and nucleotide requirements of protein–DNA interactions detected by EMSA. Another technique—methylation interference—can be coupled to the EMSA to provide information about the specific nucleotide contacts involved in the interaction. Briefly, this assay begins with an EMSA probe labeled on only one end, which, after labeling and purification, is modified on guanine bases with dimethylsulfate. For this assay to succeed, the protein-binding site generally must be at least 25 or 30 bp from the radiolabeled end, with 15–20 bp of sequence following the binding site. Thus, radiolabeled probes used for methylation interference are generally longer than those used for conventional EMSAs. The concentration of dimethylsulfate and time of incubation are chosen to result in an average of one methylguanine per probe molecule. The modified probe is incubated with extract in a standard EMSA-binding reaction. DNA-binding proteins bind randomly to the modified and unmodified probe molecules. However, a protein may be incapable of binding probe molecules in which a guanine that must be in close contact with the protein is methylated. The EMSA gel is used to separate the free probe molecules from the protein–DNA complexes. Following autoradiography, the film and gel are aligned, allowing excision of polyacrylamide gel slices containing the free probe and the protein–DNA complexes. The DNAs are eluted from the two gel slices and are then incubated with piperidine, which cleaves each probe molecule at the nucleotide(s) containing methylguanine. The resulting DNA fragments are analyzed by PAGE, followed by autoradiography or Phosphorimager analysis. If methylation of a particular guanine prevents DNA binding, DNA fragments cleaved at that guanine will be absent in the sample derived from the protein–DNA complex but will be abundant in the sample derived from the free DNA probe. Because methyl groups are quite small, the inhibition of protein binding by a particular methylguanine residue suggests that the protein is in close proximity with that guanine.

In many respects, methylation interference results can provide information that is similar to that provided by the EMSA competition and mutant studies: They can show that binding is sequence-specific and can reveal the location of the binding site. A limitation of the technique is that it provides only partial information about the nucleotides required for the protein–DNA interaction. A careful mutant analysis can provide much more detailed information, which can be correlated more effectively with the results of functional analyses. In addition, methylation interference can be technically challenging if the protein–DNA complex is not abundant. For weak complexes, it may not be possible to obtain compelling results, making the competition and mutant studies essential. For strong complexes, the use of all three approaches is recommended to provide as complete a picture as possible of the protein–DNA interaction. Additional techniques (not described here) can be used to model protein–DNA interactions more rigorously, but this detailed analysis is usually performed with recombinant proteins after the gene encoding the protein has been identified.

Analysis of Previously Described Proteins in an EMSA Complex

After the sequence specificity and nucleotide requirements for a protein–DNA interaction have been determined, it often is important to address the possibility that the protein responsible for the interaction has been described previously. The first step of this analysis is to search a binding-site database to determine whether the DNA element is similar to the recognition site for one or more known proteins. If candidate proteins are identified by the database search, the possibility that they are responsible for the EMSA complex can be evaluated using antibodies. Antibodies directed against many known DNA-binding proteins are available from a variety of commercial and academic sources. Otherwise, antibodies can be prepared against a synthetic peptide or a bacterially expressed fusion protein (see Harlow and Lane 1999).

To determine whether an EMSA complex contains a known protein, an antibody can be added directly to the binding reaction or preincubated with the extract before addition of radiolabeled probe. Preincubation may permit more efficient formation of the antibody–antigen complex if the DNA-binding domain is recognized by the antibody. In such a case, the antibody may prevent DNA binding, leading to a reduction in the amount of protein–DNA complex observed on the EMSA gel image (Fig. 2A). If the antibody recognizes a domain that is distinct from the DNA-binding domain, it is more likely to "supershift" the complex, meaning that the complex will migrate more slowly through the EMSA gel because its molecular weight will be higher when bound to the antibody (Fig. 2B). An antibody can also stabilize a protein–DNA interaction by stabilizing the protein in a binding-competent conformation. This results in a supershifted band that is more abundant than the original band observed in the absence of antibody.

Figure 2.

Antibodies are useful for determining the identity of a protein within an EMSA complex. However, they must be used with caution, primarily because most polyclonal and monoclonal antibody preparations are impure. The contaminants within the preparations can inhibit complex formation as effectively as a specific antibody–antigen interaction. Carefully designed controls must be included, particularly if the antibody inhibits, rather than supershifts, a complex. Appropriate controls for a polyclonal antibody are preimmune serum or a polyclonal preparation directed against an unrelated protein. It is also helpful to test the effect of an antibody preparation on an EMSA complex obtained with an unrelated probe bound by a protein that should not be recognized by the antibody. The specificity of the antibody effects can sometimes be greatly improved by purifying the antibodies by protein A–Sepharose or protein G–Sepharose chromatography or by antigen affinity chromatography (see Harlow and Lane 1999). If purified polyclonal antibodies are used in conjunction with the EMSA experiments, control antibodies should be subjected to the same purification protocol and analyzed along with the test antibodies. If monoclonal antibodies are used, a control monoclonal antibody prepared by the same method and directed against an unrelated protein should be included.

If the antibodies supershift rather than inhibit the complex, the controls described above are somewhat less important. However, supershift results must be interpreted with equal caution. In some instances, the abundance of a protein–DNA complex will not be diminished by the antibody, even though a supershifted complex of weak or moderate intensity appears on the gel image (Fig. 2C). This result does not establish that the protein recognized by the antibody is a component of the original complex. Because some antibodies stabilize protein–DNA interactions (see above), the supershifted complex might correspond to a stabilized form of a complex that was not detected in the absence of antibody, unrelated to the original protein–DNA complex. Evidence that a complex contains a particular protein is provided only when the abundance of the complex decreases as the supershifted complex appears. Another caveat of antibody supershifts is that antibodies can promote cooperative binding of a protein to two adjacent sites.

If the antibody does not disrupt or supershift the complex, the protein recognized by the antibody may not be present within the complex. Alternatively, the antibody could be too dilute or might bind to the target protein with a low affinity. A positive control is needed, in which the antibody is shown to disrupt or supershift an EMSA complex known to contain the protein.

If the protein–DNA complex is convincingly disrupted or supershifted by an antibody against a known protein, the complex is likely to contain that protein. To test this hypothesis further, a cDNA for the protein can be obtained and analyzed. The full-length protein can often be produced by in vitro transcription/translation (Ausubel et al. 1994, Unit 10.17) and tested by EMSA side by side with the crude cell extract. If the complex detected with the crude extract contains only one protein, it should comigrate with the EMSA complex containing the in vitro–translated protein. If not, it may be necessary to reevaluate the antibody results. Alternatively, the protein may bind as a heterodimer with another protein or may bind cooperatively with a protein that recognizes an adjacent site on the probe. Knowledge of the general properties of the protein family, and detailed knowledge of the nucleotides within the probe required for complex formation, should allow these possibilities to be evaluated.

If these studies suggest that a known protein binds the control element of interest, the relevance of the protein for the function of the element can be evaluated. In contrast, if the results suggest that the protein has not been previously described, the identity of the protein and the gene encoding the protein can be pursued. A final possibility is that the complex contains a heterodimer of a known protein and a novel partner. To identify the unknown components, the proteins that form the complex may need to be purified.