Single-Stranded DNA-Binding Proteins: Multiple Domains for Multiple Functions

Abstract

The recognition of single-stranded DNA (ssDNA) is integral to myriad cellular functions. In eukaryotes, ssDNA is present stably at the ends of chromosomes and at some promoter elements. Furthermore, it is formed transiently by several cellular processes including telomere synthesis, transcription, and DNA replication, recombination, and repair. To coordinate these diverse activities, a variety of proteins have evolved to bind ssDNA in a manner specific to their function. Here, we review the recognition of ssDNA through the analysis of high-resolution structures of proteins in complex with ssDNA. This functionally diverse set of proteins arises from a limited set of structural motifs that can be modified and arranged to achieve distinct activities, including a range of ligand specificities. We also investigate the ways in which these domains interact in the context of large multidomain proteins/complexes. These comparisons reveal the structural features that define the range of functions exhibited by these proteins.

The appropriate recognition and processing of single-stranded DNA (ssDNA) require a diverse set of binding proteins. Each instance of ssDNA must be managed appropriately and its aberrant presence recognized and resolved in an efficient manner. Inappropriate recognition or processing of ssDNA can result in chromosomal damage leading to cancer, senescence, or cell death. ssDNA-binding proteins have a wide range of structures and functions, but many of them contain small autonomous domains whose recognition of ssDNA has been well studied. These domains include four structural topologies that have been structurally characterized with ssDNA: oligonucleotide/oligosaccharide/oligopeptide-binding (OB) folds, K homology (KH) domains, RNA recognition motifs (RRMs), and whirly domains. In this review, we compare these domains and how they bind ssDNA. Additionally, we describe the way in which they achieve or avoid specificity with respect to both DNA sequence and ssRNA. Finally, we describe how these domains work together to fulfill their pleiotropic roles.

A comprehensive description of all proteins known to interact with ssDNA is outside of the scope of this review; due to space limitations, we have excluded many fascinating systems that interact with ssDNA in other contexts. We direct the reader to other sources for insights into proteins that bind both dsDNA and ssDNA (Chen et al., 2008; Duderstadt et al., 2011; Huang et al., 2012), interact with ssDNA transiently (de Silva et al., 2009; Itsathitphaisarn et al., 2012; Zhang et al., 2012), utilize double-stranded shape to aid in recognition (Barabas et al., 2008; Boer et al., 2006; Messing et al., 2012; Zhang et al., 2012), and bind only one or two nucleotides (Eastberg et al., 2004; Jaudzems et al., 2012). We have also excluded proteins that lack a high-resolution bound structure.

OB Folds

OB folds are multifunctional domains found in many areas of biology (reviewed in Murzin, 1993; Theobald et al., 2003) (Figure 1A). Because this fold is notoriously difficult to predict from primary sequence, the full representation of these domains in the proteome remains unknown. OB folds are formed from a five-stranded β barrel with interspersed loop and helical elements. They show significant structural divergence and are capable of binding a variety of ligands in addition to ssDNA and ssRNA (Theobald et al., 2003).

Figure 1. Overview of ssDNA-Binding Domains.

Proteins are colored tan, DNA is colored cyan, oxygen is colored red, nitrogen is colored blue, and sulfur is colored yellow. Regions of interest are highlighted in green.

(A) OB-A from RPA70 (Protein Data Bank [PDB] ID code 1JMC) binds ssDNA using aromatic and cation-π-stacking interactions, hydrophobic interactions, and base-mediated H bonds. Variable loop regions are shown in green.

(B) KH1 from PCBP1 (PDB ID code 3VKE) binds ssDNA with more phosphate-backbone contacts, but no intermolecular stacking interactions. The conserved GXXG motif is shown in green.

(C) RRM1 from hnRNP A1 (PDB ID code 2UP1) binds ssDNA using a variety of interactions similar to those seen in the OB folds and KH domains. Conserved RNP sequence motifs are shown in green.

(D) Two of the four subunits of the Why2 complex (PDB ID code 3N1J) are shown in tan and green. The interface has several hydrophobic and stacking interactions, but almost no base-mediated H bonds.

Known ssDNA ligands range in length from 3 to 11 nt per OB fold and bind with dissociation constants that range from low-picomolar to high-micromolar levels. Affinities roughly correlate with the length of ssDNA bound and increase with the involvement of additional domains. DNA is bound across a surface of the β barrel generally centered on strands 2 and 3 (Theobald et al., 2003). Loops augment the binding pocket, but can vary dramatically in size and participation in ligand recognition. The DNA typically binds with the bases toward the protein and the backbone solvent exposed and has an almost universally conserved polarity across the binding surface with the 5′ end closer to β3 and the 3′ end closer to β2 (Figure 1A). The nucleotide bases participate in both intra- and intermolecular aromatic stacking as well as cation-π stacking. Hydrophobic and hydrogen-bonding interactions are also common with the base and ribose moieties, and salt bridges and H bonds are formed with the phosphate backbone, although these interactions contribute less to binding.

OB folds are versatile ligand binders whose properties are tailored to their needed role with binding modes evolved to accommodate varying lengths of ssDNA and a wide range of sequence specificities and affinities. Several proteins use OB folds to bind ssDNA independent of sequence (Bochkarev et al., 1997; Raghunathan et al., 2000; Yang et al., 2002). The highly conserved eukaryotic replication protein A (RPA) plays essential roles in DNA replication, recombination, and repair by binding and protecting ssDNA while recruiting necessary cofactors (reviewed in Fanning et al., 2006). RPA is a heterotrimer made up of the subunits RPA70, RPA32, and RPA14. These subunits contain four, one, and one OB folds, respectively, and the complex binds a 25 nt ligand that contacts four of the six OB folds (Fan and Pavletich, 2012). The dual-OB-fold domain RPA70AB has also been crystallized with an 8 nt ligand that represents a second binding mode utilized by the complex (Bochkarev et al., 1997). Single-stranded DNA-binding protein (SSB) is generally thought of as a bacterial homolog to RPA, and SSBs from several bacterial species have been cocrystallized with 35 nt ligands (Antony et al., 2012; Chan et al., 2009; George et al., 2012; Raghunathan et al., 2000; Yadav et al., 2012). Most of these structures reveal homotetramers of single OB folds that all contact the ssDNA. In an elaboration on this theme, SSB from Deinococcus radiodurans also has four OB folds in a similar arrangement, but the OB folds are distributed across two subunits rather than four (George et al., 2012). BRCA2 is a multidomain protein that primarily functions in DNA recombination (reviewed in Holloman, 2011). Centrally located BRC repeats interact with Rad51, whereas two C-terminal OB folds bind ssDNA and a third OB fold is involved in protein-protein interactions (Yang et al., 2002). A large helical tower protrudes from one of the OB folds, which augments what is otherwise a canonical DNA-binding surface.

OB folds can also bind ssDNA with high sequence specificity. Telomere-end protection (TEP) proteins utilize OB folds to sequence specifically bind the GT-rich 3′ ssDNA overhang constitutively found at the end of eukaryotic telomeres (reviewed in Horvath, 2011; Lewis and Wuttke, 2012). These proteins include Pot1, Cdc13, and TEBP, which are all responsible for coordinating end protection and telomerase recruitment at the telomere. Pot1 is conserved from fission yeast to humans and binds ssDNA using a dual-OB fold module. The structure of the human POT1 dual-OB fold DNA-binding domain (DBD) was determined in complex with a 10 nt ligand, and the individual OB folds from Schizosaccharomyces pombe Pot1 have been crystallized with their minimal ssDNA ligands (Dickey et al., 2013; Lei et al., 2004). Sequence specificity is conferred by both OB folds from human POT1, but only the first OB fold from S. pombe (Altschuler et al., 2011; Lei et al., 2004). Structures of the second OB fold from S. pombe Pot1 in complex with a variety of ssDNA sequences give insight into this surprising lack of sequence specificity (Dickey et al., 2013).

The TEP protein from the organism Sterkiella nova has two protein subunits (TEBP α and β) that together use three OB folds to bind a 12 nt ligand (Horvath et al., 1998). Accommodation of alternative ligands occurs through a nucleotide-shuffling mechanism (Theobald and Schultz, 2003). TEBPα, a homolog to Pot1, also binds ssDNA on its own in a manner similar to that seen in the α-β complex (Baumann and Cech, 2001; Classen et al., 2001; Peersen et al., 2002). TEBP β shares structural similarity with the human telomerase processivity factor and binding partner of POT1, TPP1, but the functional relationship of the two complexes has yet to be reconciled (Wang et al., 2007; Xin et al., 2007). Tetrahymena Teb1 also specifically binds telomeric ssDNA and increases processivity of the telomerase enzyme (Min and Collins, 2010). Teb1 contains four OB folds, two of which are involved in binding ssDNA and have been crystallized with a 10 nt ligand (Zeng et al., 2011). The four nucleotides at the 5′ end bind in a canonical fashion, but the remaining DNA is poorly defined, possibly due to the deletion of the linker between OB folds.

In budding yeast, the TEP protein Cdc13 is part of an RPA-like complex that includes Stn1 and Ten1 (Gao et al., 2007; Gelinas et al., 2009; Sun et al., 2009). A single OB fold from Cdc13 binds an extended 11 nt ligand with picomolar affinity (Anderson et al., 2002). By analogy to the RPA complex, Stn1 and Ten1 may also contribute to ssDNA binding, but to date only the DBD from Saccharomyces cerevisiae Cdc13 has been structurally characterized with ssDNA (Mitton-Fry et al., 2002, 2004).

Sequence specificity has been observed in the binding of cold shock domains, a subfamily of OB folds characterized by a smaller, more compact topology (reviewed in Horn et al., 2007). Whereas these multifunctional domains predominantly bind mRNA to influence transcription, stability, and translation, they also bind ssDNA, playing roles in DNA repair and transcriptional regulation (Graumann and Marahiel, 1994; MacDonald et al., 1995). The structures of CspB from Bacillus caldolyticus and Bacillus subtilis in complex with 6 nt ssDNA ligands (Max et al., 2006, 2007) reveal oligomeric states facilitated by either bridging DNA or domain-swapped proteins, but the solution structure of the B. subtilis protein with the higher-affinity 7 nt poly(T) ligand suggests a monomeric conformation (Zeeb et al., 2006).

KH Domains

KH domains are small domains (~70 aa) characterized by three α helices packed against a three-stranded β sheet (reviewed in Valverde et al., 2008) (Figure 1B). Proteins that contain KH domains are significantly more common in protein databases than those with OB folds. This is likely due to their bona fide increased abundance as well as their ease of prediction from primary sequence.

KH domains typically bind a core 4 nt sequence with one or two additional nucleotides occasionally making contact with the domain, achieving an affinity in the high-nanomolar to low-micromolar range (Valverde et al., 2008; Yoga et al., 2012). The core DNA-binding pocket is formed by one β strand and two α helices connected by a loop containing a conserved GXXG motif and, like OB folds, the DNA is bound with a conserved polarity and generally oriented with the bases toward the protein. However, electrostatic interactions between the protein and phosphate backbone appear more common than in OB folds. Bases are involved in H bonding and hydrophobic interactions but, unlike the other domains described, there are no known instances of intermolecular aromatic stacking interactions.

KH domains from four different proteins have been structurally characterized in complex with ssDNA: heterogeneous ribonucleoprotein K (hnRNP K), far upstream element (FUSE)-binding protein (FBP), and poly(C)-binding proteins (PCBP) 1 and 2. Whereas most hnRNPs participate in RNA binding and processing, hnRNP K also binds the single-stranded CT element upstream of the c-myc promoter, activating transcription (Michelotti et al., 1996; Tomonaga and Levens, 1995). hnRNP K uses three KH domains to bind ssDNA, and one of these has been structurally characterized bound to three different ligands (Backe et al., 2005; Braddock et al., 2002b). c-myc transcription is further promoted by the binding of FBP to the single-stranded FUSE (Duncan et al., 1994; Liu et al., 2006). FBP contains four KH domains that all bind ssDNA, and the solution structure of a construct containing the last two (KH3+KH4) was determined in complex with a 29 nt FUSE sequence (Braddock et al., 2002a).

Similarly, PCBP (also known as hnRNP E and α-cp) 1 and 2 have been implicated in a range of functions including transcriptional activation, and their individual KH domains display ssDNA-binding activity in isolation (Choi et al., 2009; Kim et al., 2005; Yoga et al., 2012). The structures of PCBP2 KH1 and KH3 and PCBP1 KH1 in complex with ssDNA, along with the structures of the KH1 domains of PCBP1 and 2 with multiple ssDNA sequences, reveal the mechanisms defining specificity for C-rich oligonucleotides (Du et al., 2007, 2005; Fenn et al., 2007; Yoga et al., 2012). Analysis of free full-length PCBP2 in solution revealed an interaction between the first two KH domains, but not the third, suggesting potential arrangements of the full-length protein upon binding ssDNA (Du et al., 2008).

KH domains appear to be utilized for functions distinct from OB folds. As opposed to OB fold-containing proteins, all of the KH domain-containing proteins described here bind ssDNA at sites upstream of promoter regions to affect transcription. However, these proteins are also heavily implicated in posttranscriptional regulation based upon their overlapping RNA-binding activities. Moreover, hnRNP K and the PCBPs specify for an ssDNA sequence strikingly similar to the telomeric C strand, and have been suggested to play a role in telomere function (Bandiera et al., 2003; Du et al., 2005; Lacroix et al., 2000). Some, if not all, of these activities are likely biologically relevant, but more work is needed to define how these myriad functions are executed.

RRMs

RRMs are an incredibly abundant domain present in >1% of annotated human proteins (UniProt Consortium, 2012). As the name suggests, RRMs most often bind RNA, but have also been shown to bind ssDNA (reviewed in Cléry et al., 2008). RRMs are typically ~90 aa in length and form a relatively large β sheet surface (more similar to OB folds than to KH domains) packed against two α helices (Figure 1C). The majority of RRMs contain two conserved sequence motifs (RNPs) on strands 1 and 3 that form the primary nucleic acid-binding surface. Residues found elsewhere in the sheet (sometimes including an additional strand) and intervening loops also contribute to nucleic acid binding. Similar to OB folds, the ssDNA is oriented with the bases toward the protein such that they participate in stacking interactions, hydrophobic packing, and H bonding. Electrostatic interactions with the phosphate backbone contribute modestly to affinities that lie in the high-nanomolar to low-micromolar range.

Although the database is replete with RRM/RNA structures, RRMs from only three proteins have been structurally characterized in complex with ssDNA: FBP-interacting repressor (FIR), hnRNP A1, and hnRNP D (also known as Auf1). FIR (a splice variant of PUF60) is a transcriptional repressor that acts in opposition to FBP at the FUSE (Liu et al., 2000). Two of the three RRMs of FIR have been crystallized with a 27 nt sequence from the FUSE (Crichlow et al., 2008). In the structure, FIR forms a homodimer mediated by RRM2, whereas RRM1 binds ssDNA. Although the ssDNA is poorly defined, it appears to form a loop incompatible with FBP binding, suggesting a mechanism for FBP antagonism.

Like many hnRNPs, A1 and D have been implicated in a wide variety of cellular processes (reviewed in Bekenstein and Soreq, 2012; White et al., 2013). These proteins do not discriminate against ssRNA, and participate in many aspects of RNA regulation (Burd and Dreyfuss, 1994; Ishikawa et al., 1993). They are also implicated in telomere maintenance, either through direct binding of the telomeric overhang or an indirect mechanism (Batista et al., 2011; Flynn et al., 2011; Ishikawa et al., 1993; McKay and Cooke, 1992; Pont et al., 2012). Both proteins contain dual RRMs, and the crystal structure of the dual RRM from hnRNP A1 (known as UP1) was determined in complex with a dual telomeric repeat as well as derivations containing unnatural bases, giving insight into the structural basis of sequence specificity (Ding et al., 1999; Myers and Shamoo, 2004; Myers et al., 2003). Additionally, the solution structure of the second RRM from hnRNP D was determined in complex with a single telomeric repeat (Enokizono et al., 2005). Although revealing important insights into ssDNA recognition, further work is needed to determine how this activity contributes to their biological roles.

Whirly Domains

Whirly domains are a distinct structural family found almost exclusively in a small group of proteins localized to mitochondria and chloroplasts in plants (reviewed in Desveaux et al., 2005). A similar fold was found in the mammalian transcriptional regulator PurA (Graebsch et al., 2009), but the only high-resolution structures available bound to ssDNA are the mitochondrial whirly protein Why2 in complex with several different 32 nt sequences (Cappadocia et al., 2010). Similar to KH- and RRM-containing proteins, whirly proteins have been implicated in a wide variety of functions including transcriptional activation, telomere maintenance, splicing, and DNA repair (Cappadocia et al., 2012 and references within).

Whirly domains are large (~180 aa) domains that contain two roughly parallel four-stranded β sheets with interspersed helical elements (Figure 1D). Individual domains form tetramers through interaction of the helices, and these tetramers further interact to form hexamers of tetramers (Cappadocia et al., 2010, 2012). As with SSB, ssDNA is wrapped around the outside of the tetramers, with bases pointed toward the protein interface. Unlike SSB, however, the DNA takes a more direct path around the tetramer along a surface formed by the β sheets and interspersed loops. Inter- and intramolecular stacking interactions are common, but relatively few H bonds and electrostatic interactions are formed with the DNA. The tetramer binds 32 nt ligands tightly, with an affinity in the low-nanomolar range.

Specificity

Sequence Discrimination

Sequence specificity of a protein or domain can be surprisingly difficult to define. A protein does not strictly bind or not bind a particular sequence, but rather binds different sequences with a continuum of affinities. There is no cutoff for biologically relevant affinity, and thus no consensus definition of specificity exists. Different experimental techniques provide different insights into specificity, with high-throughput techniques contributing differently from individual binding assays (Campbell et al., 2012; Stormo and Zhao, 2010). Ultimately, the in vivo function of a protein must be considered, especially the potential sequences that a protein will encounter. For example, a protein that is independently localized to a position of the genome may have less stringent demands on specificity than a protein that freely diffuses through an ocean of potential nucleic acid-binding partners. Many of the details behind localization and even functions of the proteins reviewed here are unknown, and thus we review structural contributions to specificity rather than attempting to decipher its biological relevance.

ssDNA-binding proteins that function in genome maintenance, such as Why2 and the OB-fold proteins RPA, SSB, and BRCA2, necessarily bind ssDNA indiscriminately, but do so with little structural homology. Each of these proteins uses multiple domains and interacts with ssDNA largely by contacting the bases, but beyond these gross features share few commonalities. ssDNA wraps around the compact multimeric protein cores of SSB and Why2, whereas RPA maintains some conformational flexibility while binding ssDNA (Brosey et al., 2013). Whirly domains are able to limit sequence specificity by minimizing H bonds that might confer specificity, whereas OB folds utilize more complicated mechanisms discussed below. Similar to the difficulties of de novo functional prediction, there are no obvious structural indicators for sequence-nonspecific proteins.

ssDNA-binding domains can achieve high sequence specificity, and perhaps the most specific of these proteins currently known are the telomeric proteins. TEBP, Teb1, and human POT1 exhibit high specificities for their respective telomeric sequences, presumably to provide their exquisitely tailored function at the telomere (Classen et al., 2003; Gottschling and Zakian, 1986; Lei et al., 2004; Min and Collins, 2010). S. pombe Pot1 and Cdc13 provide an interesting twist to this theme in that they specifically recognize telomeric ssDNA, including accommodation of naturally occurring sequence variation. Both Cdc13 and S. pombe Pot1 recognize a key telomeric sequence in the 5′ region of the DNA but bind the 3′ region with little to no specificity, despite the importance of both regions for affinity (Altschuler et al., 2011; Eldridge et al., 2006). Although not a telomeric protein, CspB also uses an OB fold to preferentially bind poly(T) oligonucleotides with a tolerance for C at some positions (Max et al., 2006, 2007; Zeeb et al., 2006).

KH domains and RRMs can also bind ssDNA sequence specifically, but these domains generally have a smaller ligand-binding site than OB folds and thus specify for fewer positions. For example, each KH domain of hnRNP K and PCBP1 and 2 contacts just 4 or 5 nt, and only the middle two are strictly specified as cytosines (Yoga et al., 2012). The presence of cytosine at positions 1 and 4 may improve affinity slightly, but in vivo, ligands and crystal structures often have other bases at these positions (Fenn et al., 2007; Takimoto et al., 1993). The KH domains of FBP also appear to have specificity for 4 nt in vitro, but these specificities may be relaxed in the context of the full-length protein and full-length FUSE sequence (Benjamin et al., 2008). Similarly, the RRMs of FIR and hnRNP D specify 3 or 4 nt sequences, and the RRMs of hnRNP A1 combine to preferentially bind a 12 nt motif; however, this preference is also relaxed with natural ligands (Abdul-Manan et al., 1996; Benjamin et al., 2008; Burd and Dreyfuss, 1994; Cukier et al., 2010; Enokizono et al., 2005; Mazan-Mamczarz et al., 2009). The relaxation of specificities in vivo suggests these systems have a low-affinity threshold for function such that even weak binding is biologically relevant. Alternatively, localization to the site of function, increasing local concentration, may be key for binding sequences with weaker affinities.

At first glance, the way in which ssDNA-binding proteins achieve sequence specificity appears obvious. Specifically recognized ssDNA is bound in a base-inward fashion that allows the protein to form base-specific H bonds with the DNA bases. For example, the first OB fold of S. pombe Pot1 (Pot1pN) discriminates against complementary base substitutions by at least two orders of magnitude at five of the six positions (Lei et al., 2002). The crystal structure of this complex shows an abundance of H bonds formed by both the Watson-Crick and Hoogsteen faces of these bases, the pattern of which could not be recapitulated by any other base (Lei et al., 2003) (Figure 2A). Nonspecific proteins, however, bind ssDNA in the same orientation and often form what appear to be base-specific H bonds as well. In contrast to sequence-specific proteins, these H bonds do not confer significant selection for or against different bases. Studies on TEBP α/β, UP1, and Pot1pC suggest potential mechanisms to explain this lack of specificity (Dickey et al., 2013; Myers and Shamoo, 2004; Theobald and Schultz, 2003). In each of these studies, a single protein was crystallized with several nucleic acid sequences. Some of these sequences were thermodynamically accommodated, and the crystal structures provide insight into the structural mechanisms behind this accommodation. An additional base can be accommodated by TEBP by shuffling nucleotides into adjacent binding pockets (Figure 2B). This is accompanied by extrusion of a base out of its binding pocket and into solvent, thereby resetting the binding register. Conservative base modifications are tolerated by UP1 through rotation of the bases around the glycosidic bond. This reorientation maintains stacking energetics while changing the pattern of H bond donors and acceptors to best fit the binding pocket (Figure 2C). Donor/acceptor patterns can additionally be modified by rearrangement of protein side chains and water molecules or by dynamic remodeling of the interface (Bhattacharya et al., 2002). Finally, Pot1pC can accommodate base substitution with larger structural rearrangements of both the DNA and protein that result in a unique binding mode (Figure 2D).

Figure 2. Base-Mediated H Bonds Are Common at Sequence-Specific Interfaces, but Do Not Always Confer Specificity.

(A) Pot1pN (tan) binds ssDNA (cyan) with many H bonds that confer high specificity (PDB ID code 1QZH).

(B) Schematic and high-resolution structures of nucleotide shuffling by TEBP. TEBP accommodates an additional 3′ nucleotide by shuffling bases to the adjacent 5′ binding site, flipping T8 into solvent, and resetting the binding register. Cognate telomeric sequence is shown in cyan (PDB ID code 1JB7) and noncognate is shown in yellow (PDB ID code 1PH1).

(C) hnRNP A1 accommodates the unnatural base 2-aminopurine (2AP) by base flipping. 2AP (DNA and protein are shown in yellow) (PDB ID code 1U1P) maintains stacking interactions and maximizes H bonding by 180° rotation around the glycosidic bond relative to the cognate guanine (cyan) (PDB ID code 2UP1).

(D) Pot1pC binds two different ssDNA sequences—GGTTACGGT (tan and cyan) (PDB ID code 4HIK) and GGTAACGGT (green and yellow) (PDB ID code 4HIO)—in globally altered binding modes. DNA base and backbone positions differ as well as loop 2/3 of the protein backbone.

These mechanisms of accommodation suggest that more than just base-mediated hydrogen bonding is crucial for specificity. Many of these mechanisms rely upon flexibility of the ssDNA and/or protein, suggesting that rigidity of the interface may be important for achieving a high level of specificity. This may be achieved somewhat paradoxically through interactions with the sugar-phosphate backbone of the DNA. With a fixed backbone, DNA is less able to adjust in response to base substitution. This phenomenon is exemplified by the KH domains of hnRNP K and the PCBPs. All of the domains that have been crystallized with ssDNA show intimate contact between the GXXG motif and the ssDNA backbone of the highly specified cytosines. In fact, each one of these GXXG motifs contains a positively charged amino acid that contacts the phosphate backbone. These observations suggest that sequence-specific protein-ssDNA interactions are achieved through a subtle balance of intermolecular interactions and dynamics that need to take into account the ability to accommodate alternate sequences. As such, specificity cannot be reliably assigned solely based on structural information and instead requires complementary analysis.

RNA/DNA Discrimination

ssRNA and ssDNA are chemically similar, and their flexibility makes discrimination even more challenging. Proteins that bind double-stranded nucleic acids can discriminate based on helical shape, but single-stranded RNA and DNA have few conformational differences. Furthermore, the base-mediated binding mode and relative scarcity of contact with the 2′ functional group displayed by these proteins impede differentiation between RNA and DNA. Thus, many of the proteins reviewed bind RNA and DNA indiscriminately. Even proteins such as PCBP1 that have a slight preference for ssDNA over RNA still bind RNA with a potentially biologically relevant nanomolar affinity (Yeap et al., 2002; Yoga et al., 2012). This makes prediction of function based on in vitro biochemical behavior tenuous. In fact, hnRNP A1, hnRNP D, hnRNP K, FBP, FIR, PCBP1, and PCBP2 have all been assigned multiple functions involving both their ssDNA- and RNA-binding ability (Bekenstein and Soreq, 2012; Bomsztyk et al., 2004; Makeyev and Liebhaber, 2002; Page-McCaw et al., 1999; White et al., 2013; Zhang and Chen, 2012).

Some proteins, however, are able to discriminate between ssDNA and ssRNA. The basis for this discrimination is unexplored in most cases, but is best understood for the protein Pot1. TERRA is a noncoding RNA with the same sequence as the telomeric overhang, thus having the potential to act as a competitor for Pot1 binding (reviewed in Feuerhahn et al., 2010). To disfavor TERRA binding, human POT1 utilizes hydrophobic pockets near the ribose 2′ carbon to achieve an ~190-fold selection against RNA of the same sequence (Nandakumar et al., 2010). One of the OB folds from S. pombe Pot1 also discriminates >200-fold against RNA, but does so via steric exclusion of the 2′ hydroxyl as well as selective recognition of the 5-methyl group on one of the thymine bases (Lei et al., 2003). A similar discrimination between thymine and uracil is seen in the B. subtilis CspB protein (Sachs et al., 2012). Thus, a variety of mechanisms can be used, often at just a few key positions, to discriminate against RNA.

Modularity

A nearly ubiquitous feature is the use of multiple homologous domains working together to confer full activity (an exception being CspB) (Zeeb et al., 2006). These multiple domains can be tethered via covalent linkage within a single polypeptide chain or as obligate complexes (Figures 3A and 3B). Interestingly, these domains appear to always occur as repeats of the same type. There are some exceptions among RNA-binding proteins, such as the IGF2 mRNA-binding proteins, but all ssDNA-binding proteins, to our knowledge, contain only one type of ssDNA-binding domain (Letunic et al., 2012).

Figure 3. ssDNA-Binding Domains Frequently Occur in Multiple Copies.

Each domain is represented by a shape (circles, OB folds; crescents, RRMs), and separate polypeptide chains are colored differently.

(A) As is the case with SSB, single-domain proteins can come together to form obligate complexes.

(B) Additionally, single proteins themselves, such as FBP, can contain multiple homologous ssDNA-binding domains.

These domains usually work together to bind ssDNA, but structural characterization of full-length proteins and complexes can be challenging due to their size and flexibility. Thus, many of these proteins have been studied using a divide-and-conquer approach in which individual domains are structurally characterized in isolation. This approach provides valuable baseline information, but further investigation is required to determine the effects of domains upon one another.

The addition of a second DNA-binding domain has the simplistic effect of increasing affinity by extending the binding interface, but there are more intricate effects as well. For example, multiple domains may also allow handoff or sliding, as seen in single-molecule experiments on SSB and the POT1-TPP1 complex (Hwang et al., 2012; Zhou et al., 2011). In SSB, one or more OB folds remain bound to ssDNA while another moves to a new binding site, thus necessitating multiple domains (Figure 4A). Although not yet described in a specific system, a similar mechanism could be used to “hand off” an ssDNA ligand between two proteins. This could occur by removing and attaching one domain at a time, again necessitating multiple domains (Figure 4B).

Figure 4. Multiple Domains Can Be Used for a Variety of Purposes.

(A) As with SSB, DNA sliding can be achieved by keeping one or more domains bound while another domain repositions itself on the ssDNA ligand.

(B) Handoff of an ssDNA ligand could be achieved by sequential dissociation and association of domains.

The length of the linker between these domains may also play an important functional role by modulating affinity, specificity, and/or conformational flexibility. A shorter linker usually results in a more additive affinity with an upper bound of the product of the two individual affinities (Shamoo et al., 1995; Zhou, 2001). A longer linker theoretically reduces the additive effect between domains to a combined affinity just greater than that of the single domain with the highest affinity. Linker length could also impact specificity. Whereas a short linker may only allow the recognition of adjacent binding sites, an extended linker might allow two domains to accommodate nucleotide spacers of various lengths (Figure 5A). Another potential function for extended linkers is to allow two domains to bind perfectly adjacent binding sites (Figure 5B). In the crystal structures of some KH domains, two domains bind adjacent sites with no nucleotide spacers by bending or rotating the DNA to avoid steric clash between protein domains (Backe et al., 2005; Yoga et al., 2012). Although the KH domains in these structures are not physically connected, an extended linker (present in many of the full-length proteins) would allow for this conformation. These, and other, functions are also seen in modular RNA-binding proteins, as expertly reviewed (Barraud and Allain, 2013; Lunde et al., 2007; Mackereth and Sattler, 2012).

Figure 5. Polypeptide Linkers Often Connect Domains within a Protein.

(A) These linkers may reduce specificity of a protein by accommodating spacers of different lengths between two binding sites, as seen in S. pombe Pot1.

(B) These linkers may also allow conformational flexibility necessary to bind adjacent ssDNA-binding sites, as seen in the structures of individual KH domains.

In many cases, the additional domains function to mediate higher-order complex formation. Cdc13, Pot1, FIR, RPA, BRCA2, and TEBP all function in protein complexes mediated by respective OB folds or RRMs (Bochkareva et al., 2002; Crichlow et al., 2008; Horvath et al., 1998; Mitchell et al., 2010; Sun et al., 2011; Wang et al., 2007; Yang et al., 2002). Some of these domains are able to bind ssDNA in addition to their role as protein-interaction domains, whereas others have lost their ability to bind ssDNA. Thus, “ssDNA-binding” domains can function as protein-protein interaction modules, but this ability does not exclude additional biochemical activities.

Functional Modulation

Many of the ssDNA-binding proteins described here are proposed to have multiple functions. Although in vitro structural and biochemical studies are valuable starting points for defining the function of nucleic acid-binding proteins, in most cases it is unclear what modulates these functions. One hypothesis is that the intrinsic activity of a given domain is modulated by its protein-binding partners (Figure 6A). For example, RPA can bind ssDNA using two binding modes: one that uses two OB folds and another that uses four OB folds (Brosey et al., 2013). Many protein cofactors have been suggested to influence which binding mode is used by RPA (or whether RPA is bound to ssDNA at all), thereby directing its function (e.g., Jiang et al., 2006).

Figure 6. Many ssDNA-Binding Proteins Have Multiple Functions that Can Be Regulated in a Variety of Ways.

(A) Protein-protein interactions can affect ssDNA-binding ability.

(B) Posttranslational modifications can affect protein-protein interactions and/or cellular localization.

(C) Alternative splicing, illustrated here by a reduction in linker length, can affect protein-protein interactions, posttranslational modification, and/or cellular localization.

(D) ssDNA sequence, length, or shape may affect function in as-yet unknown ways.

In addition to protein cofactors, posttranslational modifications may trigger a functional switch (Figure 6B). For example, TEBP can bind telomeric ssDNA as a dimeric α₂ complex or as an α-β complex (Peersen et al., 2002). In combination with other data, this has led to the proposal that the α₂ complex links chromosomal ends together, whereas the α-β complex is involved in capping and protecting the telomere. Phosphorylation of the β subunit was seen to disrupt the α-β complex, presumably promoting the α₂ complex and thus acting as a functional switch (Paeschke et al., 2005).

Cellular localization likely assists in dictating the function of these pleiotropic proteins (Figures 6B and 6C). For example, HIV-1 infection triggers cytoplasmic accumulation of hnRNP A1, which in turn promotes translation at an internal ribosome entry site of the viral RNA (Monette et al., 2009). Alternative splicing can also influence function, as illustrated by hnRNP D (Figure 6C). hnRNP D has four splice variants that differentially affect mRNA stability (Raineri et al., 2004; Sarkar et al., 2003). All four isoforms bind RNA, but in different conformational states (Zucconi et al., 2010). These conformational states may differentially recruit additional protein factors that ultimately affect mRNA stability. Furthermore, alternative splicing can affect posttranslational modification, which can in turn affect cellular localization (Arao et al., 2000; van der Houven van Oordt et al., 2000; Wilson et al., 2003). Thus, these mechanisms of functional control are largely intertwined.

In addition to protein-binding partners, nucleic acid sequence, length, and/or structure may influence function (Figure 6D). For example, S. pombe Pot1 uses different binding modes to bind telomeric DNA of different lengths (Altschuler et al., 2011). Although speculative, it is possible that these different binding modes act as a measure of telomere length and affect telomerase recruitment accordingly. Overall, the modulation of the intrinsic activities of these domains is a largely unexplored area of biology, and further efforts are needed to define the mechanisms of functional modulation for these, and other, pleiotropic proteins.