DNA–protein binding occurs by many modes and mechanisms
Nonspecific binding of proteins to DNA can proceed very simply through electrostatic interaction between basic groups on the protein and the smooth, regular track of negatively charged phosphates that form the phosphodiester backbone (see Figure 1 and Figure 2). In fact, virtually any positively charged protein will tend to stick to DNA in solution, at least at low ionic strength. This is often also true for proteins that possess a recognition site for a particular nucleotide sequence: they can cling to the DNA elsewhere but more weakly. This is often a factor in facilitating the access to specific protein-binding sites. Sometimes nonspecific binding is used to coat the DNA, thereby protecting it against unwanted interactions, protecting it from degradation, or compacting it. Some proteins can do all of these things. An important example is found in chromatin, the protein–DNA complex in which DNA is compacted and sequestered in the eukaryotic nucleus. Much more is said about chromatin in Chapter 8; for now, suffice it to note that in chromatin, complexes of eight basic protein molecules form spools upon which nuclear DNA can be wound. To a large extent this binding is nonspecific, as must be expected for an interaction that involves most of the genome. Nevertheless, some elements of nucleotide sequence preference or avoidance for the formation of such structure can be found.
Fig1. A tetranucleotide. The backbone of the chain is formed by the phosphates and sugars, marked in alternating blue and red for the successive nucleotide residues; bases are indicated by the circled letters. Note that the chain possesses polarity: the 5′-end and the 3′-end are chemically different. Reading the sequence from the top, in the 5′ → 3′ direction, will give ATCG; reading from the bottom, in the 3′ → 5′ direction, will give GCTA. By convention, nucleotide sequences in nucleic acids are written and read from 5′ to 3′.
Fig2. Schematic of the double-helix structure of B-form DNA. Base pairing occurs between a purine on one strand and the corresponding pyrimidine on the other strand; the pairing between these complementary bases allows the distances between the C-1′ on the sugar moieties to be exactly the same for both adenine–thymine and guanine–cytosine base pairings. Hydrogen-bonding between the complementary bases and stacking interactions between successive pairs are both important in stabilizing the helix. The original Watson Crick model closely resembles this but with 10.0 base pairs/turn. [Adapted, courtesy of the U.S. National Library of Medicine.]
There is a class of nonspecific DNA-binding proteins that preferentially interact with and stabilize single-stranded DNA (ssDNA). These single-strand binding proteins (SSB), or helix-destabilizing proteins as they are sometimes called, are involved in processes such as genetic recombination, DNA replication, and DNA repair, where it is essential to maintain a region of denatured DNA for some time. SSB proteins are ubiquitous and can function as monomers, as seen in viruses; as homotetramers, as seen in bacteria; or as heterotrimers, such as replication protein A in eukaryotes. It was long believed that such proteins might bind to a transient opening in the helix and then bind cooperatively in a side-by-side fashion to expand the opening. However, recent work in the von Hippel laboratory indicates that this cannot be a universal mode; binding sites for the SSB are often so large that spontaneous openings of sufficient size would be very rare. This kinetic block means that a helicase molecule is needed to first unwind a suitable length of DNA. The SSBs can then bind and stabilize the single strand. A well-studied example, the SSB from Escherichia coli, is illustrated in Figure 3. In vitro, the protein binds to ssDNA either as a dimer or as a tetramer, depending on the conditions. In the cell, SSB binds as a tetramer, wrap ping ~70 nucleotides of ssDNA around the protein subunits. This wrapping leads to an overall reduction in the contour length of the DNA, as seen in the electron micro scope images presented in Figure 3C.
Fig3. E. coli single-strand binding protein. SSB binds single-stranded regions of DNA to prevent premature reannealing of the two strands during DNA replication and related processes. It also protects ssDNA from nucleolytic cleavage. E. coli SSB is composed of four identical 19 kDa subunits that can interact with ssDNA in different modes depending on the environment. In the (SSB)65 binding mode depicted, ~65 nucleotides of DNA wrap around the SSB tetramer and contact all four subunits; this mode is favored at high salt concentrations. At lower salt concentrations, the (SSB)35 binding mode prevails, with ~35 nucleotides binding to only two of the SSB subunits. (A) Crystal structure of (SSB)65 depicting 70 nucleotides of ssDNA, shown as a black line, wrapped around the four SSB subunits. The protein contains a characteristic tertiary structure motif, termed oligonucleotide/oligosaccharide-binding or OB-fold, that is present in many DNA- or RNA-binding proteins; its general topology is shown in the box. (B) Cartoon representing ssDNA as a yellow ribbon wrapped around the SSB core, corresponding to the structural model in part A, with the addition of the unstructured C-terminal tails, shown as gray lines, that are not observed in the crystal structure. The nine-amino-acid sequence, shown in single-letter amino acid code, at each C-terminus is responsible for the interaction of SSB with other metabolic proteins. (C) Electron microscopic image of naked DNA (left) and SSB-bound DNA (right). The reduction in contour length is due to the wrapping of the DNA around SSB. [A and B, adapted from Kozlov AG, Jezewska MJ, Bujalowski W et al. (2010) Biochemistry 49: 3555–3566. With permission from American Chemical Society. A, inset, from Agrawal V & Kishan RKV (2001) BMC Struct Biol 1: 5, doi: 10.1186/1472-6807-1-5. With permission from Springer Science and Business Media. C, courtesy of Maria Schnos, University of Wisconsin, Madison, Wisconsin, USA.]
Site-specific binding is the most widely used mode
The binding of proteins to specific sites on DNA is fundamental to a vast range of functions and structures. It has been estimated that the human proteome contains several thousand site-specific DNA-binding proteins, each with a specific base sequence that it recognizes and binds to. Some binding proteins recognize only a single site in the whole genome, whereas others bind at multiple locations. The most intriguing question is, how do protein molecules distinguish particular DNA sequences? Even before we had specific knowledge, it was clear that the B-form double helix of DNA contained information that could be read by other molecules. The surface of the DNA duplex is defined by two deep grooves, known as the major and minor grooves (see Figure 2). The edges of the bases are exposed in each of these grooves in a way that presents a unique combination of chemical groups, which can interact with a protein that inserts into the groove. The combination includes, for example, both hydrogen-bond donors and acceptors, and these are presented in a different pattern for a GC base pair than for an AT base pair (Figure 4). Thus a protein, with its own hydrogen-bond acceptors and donors, can detect the difference between these two kinds of pairs. In addition to this, there are other kinds of interactions, such as van der Waals interactions between methyl groups on the DNA and nonpolar groups in the protein, that can convey the identity of a base pair.
Fig4. Proteins recognize specific sequences on the DNA by predominantly interacting with the major groove. Looking down the DNA helix, one can see the edges of the bases protruding into the major groove. Accessibility of these bases depends primarily on the width of the groove. Clearly, the minor groove is less accessible. A stack of several base pairs along the chain will create a unique constellation of chemical groups into the groove, which is recognized by DNA sequence-specific binding proteins.
It should be emphasized that many site-specific binders also exhibit a much weaker nonspecific binding mode, involving electrostatic interactions between positive amino acids on proteins and the featureless track of negative charges on the DNA sur face. This can help a protein find its specific site, either by sliding along the DNA helix or by hopping from one region to another across DNA loops. Such one-dimensional hunting can be more efficient than wandering through three-dimensional space.
Most recognition sites fall into a limited number of classes
There are many possible specific amino acid–nucleotide interactions that could pro mote protein recognition of a binding site (Figure 5). Analysis of many X-ray diffraction studies of protein–DNA complexes reveals no simple set of rules for such pairings. There is some preference for arginine to pair with G in GC pairs and for glutamine or asparagine to pair with AT pairs (Figure 5). Evolution appears to have adopted whatever works, often complicated by distortion of the DNA and/or protein to provide the best fit.
Fig5. Possible patterns of hydrogen-bonding during recognition of base pairs by amino acid side chains. (A) Recognition of GC base pairs by arginine, lysine, glutamine, and serine side chains. (B) Recognition of AT base pairs by glutamine side chains. Note the two possible patterns of hydrogen-bonding.
The number of base pairs involved in different DNA recognition sites varies widely, depending on the function of the protein. Some restriction endonucleases, for example, have relatively short recognition sites, 4–6 base pairs (bp), so that they can cleave genomic DNA in many places. Some transcription activator or repressor proteins, which are targeted against only one or a few sites in the genome, must have larger recognition sites. They need not be excessively large in order to make unwanted accidental false recognition unlikely. Since the number of possible combinations of bases in a site of size n is 4n, sites 10 bp long will occur only about once every million base pairs, giving an expectancy of several thousand such sites in the whole human genome of 3.2 billion bp; some of these sites may be obscured by other proteins bound to or around them. When n = 20, an astronomical 1012 different sequences are possible; thus a particular, specific 20 bp site would be unlikely to be found anywhere within the genome.
Some recognition sites, such as those of the Lac repressor binding protein (Figure 6), are palindromic and can therefore accept a dimer of the protein, increasing both the specificity and binding affinity. As Figure 6 shows, Lac repressor actually takes this one step further by recognizing two adjacent palindromic pairs and binding as a tetramer. Indeed, as we shall see in further examples, site-specific binding proteins frequently exhibit tandemly repeated binding domains, which then interact with tandemly repeated DNA sites. This mechanism can make it possible to attain high specificity and binding strength, even if there are only a few specific inter actions between each domain and the DNA.
Fig6. Binding of Lac repressor to DNA. The repressor binds to DNA via helix–turn–helix (HTH) motifs in the DNA-binding domains of two monomers that form a dimer. Note that the two half-sites in the DNA recognition sequence form a palindrome, which matches the symmetry of the dimer. The repressor binds to DNA as a tetramer, or dimer of dimers, formed by interactions of the C-termini; each dimer binds two half-sites. The monomers in the top dimer are presented in green and pink; the monomers in the bottom dimer are in violet. The Lac repressor kinks the DNA at a central CpG base pair step. The helix axes on both sides of the kink, shown in blue, show an abrupt change in the helix trajectory caused by the kink. [Right, from Wilson CJ, Zhan H, Swint Kruse L et al. (2007) Cell Mol Life Sci 64: 3–16. With permission from Springer Science and Business Media. Left, from Rohs R, Jin X, West SM et al. (2010) Annu Rev Biochem 79: 233–269. With permission from Annual Reviews.]
Finally, it should be emphasized that distortion of the binding partners, especially bending of the DNA, is a common feature of site-specific binding. Because DNA bending exacts a free energy price, this must be returned by a favorable free energy gain from the interaction process itself. In other words, the binding must be strong to compensate for the bending.
Most specific binding requires the insertion of protein into a DNA groove
DNA binding can gain stability from protein interaction with the uniform phosphate backbone of DNA. The recognition of specific sites along the DNA chain, however, demands the insertion of a portion of the protein into one of the grooves in the DNA helix, where the specific base pairs can be sensed. Most DNA-binding proteins use the major groove. This is understandable because the major groove provides more recognition sites and in B-DNA it is significantly wider than the minor groove. The major groove is, in fact, wide enough to accommodate snugly a segment of α-helix, and this is a very common mode of binding. Although less common, there are examples in which two-stranded β-sheets, or the edges of even larger β-sheets, are inserted into the major groove. The good fit of α-helix or β-sheet into the major groove means that these protein domains can bind to recognition sites with little distortion of the DNA.
Binding into the minor groove is a different matter. In the first place, the minor groove of B-DNA is occupied by a spine of precisely positioned water molecules. In consequence, binding to the minor groove involves the release of this structured water, with a large entropy increase. This appears to be a major free energy source for minor groove binding. Second, the minor groove is just too narrow to accommodate an element such as an α-helix, so if such binding is to occur, the groove must be stretched open. There are a number of ways in which DNA can accommodate such stretching, but the most common is by compression of the adjacent major groove. This will have the consequence of bending the DNA helix at the site of protein binding (Figure 7). Thus, it is not surprising that minor-groove binders are almost synonymous with DNA benders. The energetic cost of such bending appears to be largely paid by the above mentioned entropy increase resulting from the release of water. However, there may be an interesting synergy here, in that the water spine contributes to the rigidity of the double helix; so displacing it should make DNA more flexible.
Fig7. Minor-groove binders cause DNA bending. (A) DNA-binding domain of Lac repressor bound to nonspecific (top) and specific (bottom) nucleotide sequences. The hinge region, shown in red, is unstructured in both the free protein and the protein bound to nonspecific DNA; it folds into an α-helix that interacts with the minor groove in the specific complex. In the nonspecific complex the DNA adopts the canonical B-form conformation, whereas in the specific complex it is bent by ~36°. (B) Crystal structure of the integration host factor (IHF) complex with DNA: α subunit is shown in cyan, β subunit in purple, consensus sequence DNA in green, and less conserved DNA in blue. IHF is a small, 20 kDa heterodimeric protein that binds DNA in a sequence specific manner and induces a large bend of >160°. This bending aids in the formation of higher-order structures in such processes as recombination, trans position, replication, and transcription. The protein contacts only the phosphodiester backbone and some bases in the minor groove; thus, it represents an example of indirect readout, where the protein relies on sequence-dependent structural features of the DNA, such as backbone conformation and flexibility. This sequence recognition is in contrast with direct readout, where sequences are distinguished through the unique functional groups of DNA bases in the major groove. [A, from Kalodimos CG, Biris N, Bonvin AMJJ et al. (2004) Science 305: 386–389. With permis sion from American Association for the Advancement of Science. B, from Lynch TW, Read EK, Mattis AN et al. (2003) J Mol Biol 330: 493–502. With permission from Elsevier.]
It should be emphasized that major-groove binding can also give rise to bending. This can happen when a dimeric or oligomeric protein binds to adjacent sites in the DNA. There exist a number of ingenious ways to study DNA bending and relate it to protein binding.
Some proteins cause DNA looping
We have noted that many DNA-binding proteins exist as noncovalently-bound dimers or even larger oligomers. If such proteins have recognition sites on DNA that are some distance apart, they have the capability to produce DNA loops (Figure 8).
Fig8. DNA looping. (A) Three-dimensional view of a loop formed by binding of a tetramer of the Lac repressor from E. coli to the two operators that control the expression of the lac operon. The tetramer is in dark blue; the DNA is represented by a combination of stick and space-filling models, with the backbone in red and purple. (B) Physical models corresponding to possible loop geometries. Photographs show loop models with the V-shaped or crystallographic repressor conformation (left column) and the corresponding configurations that result when the repressor is opened (right column). In each pair, the structure on the right was produced from that on the left by rotating the two half-tetramers, represented by a blue or a red clip, away from each other about the axis of the four-helix bundle, represented by a silver bolt. The paper strip representing the DNA is colored black on one side and white on the other to make any twist in the helix visible. [A, courtesy of Elizabeth Villa, Hughes Medical Institute, University of California, San Diego, CA, USA. B, from Wong OK, Guthold M, Erie DA et al. (2008) PLoS Biol 6: doi: 10.1371/journal .pbio .00602 32.]
DNA looping is observed in both prokaryotic and eukaryotic genomes. In many cases, it appears to isolate particular regions of the genome so as to regulate transcription therein.
There are a few major protein motifs of DNA-binding domains
Every protein that binds specifically to DNA for a specific function needs at least two domains. One is called the transactivating domain. It either senses an external signal to prompt the protein to bind, as in a transcription factor, for example, or it will carry out some process on the DNA, as a restriction endonuclease, for another example. The other domain is the one we have been concerned with so far: the binding domain that attaches to the DNA at a specific site or sites. There may be multiple transactivating domains, or complexes of several proteins that are noncovalently attached to the binding domain. In many cases, the transactivating domains can cause con formational changes in the binding domains. This may suggest that there should be caution about the many experiments, especially X-ray structure determinations, in which only the purified or cloned DNA-binding domain, bound to DNA, is studied. This structure might not reflect the structure of the entire bound protein.
Surprisingly, a large fraction of binding domains appears to use only a small vocabulary of protein motifs, known as recognition motifs, for the actual binding, even if the binding serves very different functions. This may suggest that the DNA-binding domains of proteins evolved from only a few examples, ages ago. We describe here just three of the most frequently encountered motifs; they are depicted schematically in Figure 9.
Fig9. DNA-binding motifs. (A) Helix–turn–helix, HTH; (B) zinc finger; (C) leucine zipper.
Helix–turn–helix motif interacts with the major groove
Helix–turn–helix (HTH) motifs are common in both prokaryotic and eukaryotic transcription factors. The motif consists of a stretch of ~20 amino acid residues, divided into two α-helices, each about 7–8 residues long, separated by a turn or loop. The second of these two helices, from the N-terminus, is the recognition helix, which lies in the major groove of the DNA. A specific example, CRP activator, is shown in Figure 10. Here there are three α-helices. As the α-helix has 3.6 residues per turn, only two turns are being presented to the DNA, permitting the opportunity for only a few specific contacts. In many cases, HTH proteins bind as dimers or even tetramers, increasing the number of contacts and hence selectivity (see Figure 10). As mentioned above, the binding of more than one monomer can have another consequence: bending of the DNA. As Figure 10 illustrates, the geometry of protein–protein interactions can make it necessary for the DNA to bend in order to interact with all of the HTH motifs.
Fig10. Helix–turn–helix protein bound to DNA. CRP activator is a cyclic AMP regulatory protein, also known as CAP or catabolite activator protein, that positively controls numerous operons in E. coli. Its activity is regulated by cyclic AMP binding. The protein binds as a dimer to a bipartite binding site, shown in green and blue, causing the DNA to bend ~90°. Each monomer has two structural domains: the N-terminal domain, amino acids 1–140, contains the cyclic AMP nucleotide binding site, whereas the C-terminal 50–60 amino acid domain contains a helix–turn–helix motif that interacts with the DNA.
Zinc fingers also probe the major groove
Another motif frequently observed in DNA-binding domains is called the zinc finger because of the essential role of zinc in its structure. As shown in Figure 11, the most common variant has a short length of α-helix bound to a short β-sheet via the zinc atom. Usually, the zinc is coordinated to two histidine side chains on the α-helix and two sulfhydryl groups of cysteine residues on the β-sheet, but sometimes four sulfhydryls are used. In any event, the compact finger that is produced can be inserted into the major groove without distortion. The number of possible interactions is small, so specificity is often gained by having multiple fingers. These may be all contained in the sequence of one polypeptide chain, as in the TFIIIA transcription factor shown in Figure 12, or they may be contained in interacting proteins, as in the steroid receptor proteins in Figure 13.
Fig11. Some proteins bind DNA with multiple zinc fingers. (A) One of the three zinc fingers of transcription factor Zif268. In this case, the Zn atom is coordinated between two cysteine and two histidine residues. (B) The middle zinc finger of Zif268 bound to DNA, shown in purple, in the major groove, with the Zn(II) atom shown as a sphere. [A, courtesy of Thomas Splettstoesser, Wikimedia. B, from Magliery TJ & Regan L (2005) BMC Bioinf 6: doi: 10.1186/1471-2105-6-240. With permission from BioMed Central.]
Fig12. Arrangement of proteins containing multiple zinc fingers along the DNA. (A) Transcription factor TFIIIA binds to 5S rDNA, the gene that encodes the 5S RNA component of the ribosome, via multiple zinc fingers (ZF) that insert themselves into the major groove. The two major recognition regions, blocks A and C, are contacted by fingers 7–9 and 1–3, respectively. Fingers 4–6 are used when the protein interacts with RNA. (B) Crystal structure of fingers 1–3 bound to DNA. [A, adapted from Dyson HJ (2012) Mol BioSyst 8: 97–104. With permission from Royal Society of Chemistry, UK. Originally from Clemens KR, Liao X, Wolf V et al. (1992) Proc Nat Acad Sci 89:10822–10827. Used with permission.]
Fig13. Zinc fingers in steroid receptors. (A) The sequences in DNA that bind steroid hormone receptors, called steroid response elements, consist of two short half-sites that may be palindromic or direct repeats. The receptor proteins bind palindromic sequences as head-to-head homodimers; receptor protein heterodimers bind direct repeats. The schematic shows examples of each binding mode. (B) Crystal structure of the estrogen receptor bound to DNA. The monomer receptors have two zinc fingers each. The binding of the first finger determines the sequence-specific binding, while the second finger is responsible for dimerization; that is, zinc fingers can also be protein–protein interaction domains.
Leucine zippers are especially suited for dimeric sites
Leucine zippers are protein–protein interaction motifs that help the stable formation of protein dimers to increase the specificity of interaction with DNA. Leucine zipper proteins always interact with DNA as homo- or heterodimers, held together by hydrophobic interactions of long α-helices (Figure 14A). These interacting protein tails each have leucine or isoleucine residues spaced about 3–4 residues apart. This means that all of these residues will lie on one side of the α-helix, creating a hydro phobic face (Figure 14B). This face can be buried away from solvent by having the helices make a gentle coiled coil about one another. The term zipper refers to the fact that isoleucine residues on the two α-helices often interdigitate like the teeth of a zipper. At the end of this coil lie the recognition elements themselves, usually α-helix segments that insert into the major groove (Figure 14C). Zippers can be homodimers, which interact with a repeated DNA sequence, or heterodimers, which carry different DNA-recognition elements and thus interact with a pair of different DNA sites. This allows a sophisticated kind of control, in which two different protein factors must be present and interact to allow DNA binding.
Fig14. Leucine zippers. Leucine zippers are protein–protein interaction motifs that help the stable formation of protein dimers to increase the specificity of interaction with DNA. (A) Leucine zippers form via side-by-side interactions of amphipathic helices. These are characterized by the presence of hydrophobic, nonpolar residues on one side of the helical cylinder and hydrophilic and polar residues on the other side, resulting in different properties of the two sides. The colors of the side chains represent their hydrophobicity ranging from hydrophilic, shown in blue, to hydrophobic, shown in red. The leucine residues on one side of the helix interdigitate with the leucine residues on another helix from another protein molecule. (B) Amphipathic helices can be predicted by helical wheel analysis, in which a stretch of amino acid sequence is imaginarily arranged in a wheel that mimics the arrangement of amino acids in an α-helix. The example given is that of 20 N-terminal amino acid residues of chicken cathelicidin, chCATH-B1. Note that, in addition to leucines, the hydrophobic top side of the helix also contains isoleucine, another hydrophobic amino acid residue. (C) Structure of a leucine zipper protein heterodimer bound to DNA. The DNA-binding domains are usually basic. [A, courtesy of David E. Volk, University of Texas Health Science Center, Texas, USA. B, adapted from Goitsuka R, Chen CH, Benyon L et al. (2007) Proc Natl Acad Sci USA 104: 15063–15068. With permission from National Academy of Sciences. Copyright (2007) National Academy of Sciences, USA.]
The DNA-binding elements of proteins described in this section do not represent anything like the totality of modes of protein–DNA interactions. There are many variants on the motifs described above, and there are many proteins that seem to bind in unique ways. Important examples of protein–DNA complexes feature throughout this book.
