Single-stranded telomere-binding protein employs a dual rheostat for binding affinity and specificity that drives function.

ssDNA, which is involved in numerous aspects of chromosome biology, is managed by a suite of proteins with tailored activities. The majority of these proteins bind ssDNA indiscriminately, exhibiting little apparent sequence preference. However, there are several notable exceptions, including the Saccharomyces cerevisiae Cdc13 protein, which is vital for yeast telomere maintenance. Cdc13 is one of the tightest known binders of ssDNA and is specific for G-rich telomeric sequences. To investigate how these two different biochemical features, affinity and specificity, contribute to function, we created an unbiased panel of alanine mutations across the Cdc13 DNA-binding interface, including several aromatic amino acids that play critical roles in binding activity. A subset of mutant proteins exhibited significant loss in affinity in vitro that, as expected, conferred a profound loss of viability in vivo. Unexpectedly, a second category of mutant proteins displayed an increase in specificity, manifested as an inability to accommodate changes in ssDNA sequence. Yeast strains with specificity-enhanced mutations displayed a gradient of viability in vivo that paralleled the loss in sequence tolerance in vitro, arguing that binding specificity can be fine-tuned to ensure optimal function. We propose that DNA binding by Cdc13 employs a highly cooperative interface whereby sequence diversity is accommodated through plastic binding modes. This suggests that sequence specificity is not a binary choice but rather is a continuum. Even in proteins that are thought to be specific nucleic acid binders, sequence tolerance through the utilization of multiple binding modes may be a broader phenomenon than previously appreciated.

The proper management of ssDNA in the cell is required for numerous aspects of chromosome biology. In all kingdoms of life, ssDNA is formed transiently during the execution of many essential cellular processes including transcription, DNA replication, recombination, and repair. To coordinate these numerous activities, a diverse array of proteins has evolved to bind ssDNA, to facilitate normal events such as DNA replication, or to signal the appearance of inappropriate ssDNA and initiate repair (1).

Several of these ssDNA-binding proteins function in genome-wide maintenance (1, 2). Widely studied examples include the bacterial single-strand–binding protein (SSB) and its functional equivalent in eukaryotes, replication protein A (RPA) (3–5). SSB and RPA are both essential for DNA replication, binding nascent ssDNA that is generated when duplex DNA is unwound and thereby preventing reannealing and/or the formation of secondary structures that would impede progression of the replisome. Both proteins are also central to the cellular response to DNA lesions. Although RPA and SSB exhibit no sequence homology, they each employ an array of OB-folds for contacting ssDNA. Detailed analysis of RPA has revealed that it utilizes these OB-folds to contact ssDNA in distinct modes, engaging differing lengths of ssDNA with different subunits, presumably to orchestrate higher-order manipulations (6–8). Thus, to interact consistently throughout the genome, RPA also needs to bind ssDNA indiscriminately. Commensurate with this expectation, RPA displays little obvious sequence preference in vitro, binding ssDNA tenaciously with single-digit nanomolar affinities (6, 7).

In contrast to the genome-wide and the apparently sequence-nonspecific role performed by the canonical RPA complex (2), proteins that interact with ssDNA overhangs at telomeres exhibit sequence specificity tuned to the G-rich telomeric repeats (9–11). These telomere-dedicated proteins also show exceptional affinities for their ssDNA ligands, ranging from the tight nanomolar binding by human Pot1 to single-digit picomolar binding by the Saccharomyces cerevisiae Cdc13 protein (9, 11, 12). Remarkably, the Cdc13 protein performs its telomere-dedicated role as a subunit of a heterotrimeric complex with a domain architecture that closely parallels that of RPA (13). In both the canonical and telomere-dedicated RPA, the large subunit is constitutively associated with two smaller proteins, Stn1/Ten1 with Cdc13 and Rpa32/Rpa14 with RPA70. In the CST complex, high affinity for ssDNA is conferred by the large subunit, whereas in RPA high affinity is achieved through multivalency (2, 4, 8, 12–15). However, ssDNA binding by the telomere-dedicated RPA complex (t-RPA) is notably distinct from RPA, suggesting that these structurally similar domains have taken on distinct biochemical roles. Unlike RPA70, which uses two OB-fold domains for its core recognition of ssDNA, Cdc13 employs a single OB-fold augmented by an unusually long β2–3 loop (Fig. 1) (16) to contact DNA with exceptionally tight picomolar affinity. Furthermore, Cdc13 binds ssDNA with exquisite specificity for G-rich sequences (17, 18), which it achieves through recognition of a GxGT motif embedded in a larger oligonucleotide (10, 12). Nevertheless, Cdc13 presumably needs to show sequence flexibility to accommodate the heterogeneity of yeast telomeres (19), although the mechanism by which Cdc13 achieves this flexibility has not previously been elucidated.

Fig. 1.

Mutations across the surface of the Cdc13 DBD exhibit a range of loss of affinity. (A) Structure of the Cdc13 DBD/Tel11 complex with sites of interest highlighted. The backbone of the protein is shown in wheat, with amino acids that have been individually substituted with alanine highlighted in red, purple, and blue. The β2–3 loop is shown in blue. Tel11 is in gray and green, with green bonds highlighting sites modified for the investigation of specificity (30). (B) Bar graphs show the fold loss of binding affinities of WT and mutant Cdc13 DBD for the Tel11 substrate with associated error. The Inset highlights the six mutants with modest changes in affinity on an expanded scale.

The sequence specificity and affinity displayed by Cdc13 provide a unique system for investigating how these two biochemical properties contribute to function in vivo. To do so, this study examined an extensive panel of mutations across the DNA-binding interface for their effects on both binding affinity and specificity and subsequently determined how perturbations in either property affected Cdc13 function in vivo. Not surprisingly, substantial reductions in Cdc13-binding affinity were lethal in vivo, whereas less severe declines in affinity were better tolerated. Unexpectedly, this approach also identified a second category of mutations that had little effect on affinity but large effects on specificity in vitro. These specificity mutations reduced the ability of Cdc13 to tolerate variations in telomere sequence, which substantially impaired Cdc13 function in vivo. Moreover, the magnitude of the in vivo defect closely correlated with the extent to which specificity was altered, thereby demonstrating that both affinity and sequence tolerance contribute to biological function. Thus, by conducting a systematic analysis of the Cdc13 DNA-binding domain (DBD) interface, we have uncovered a finely tuned binding rheostat of specificity and affinity that confers optimal biological function.

Results

Systematic Mutagenesis of the DNA-Binding Interface of Cdc13 Identifies a 35-Fold Span in Affinity.

To address how the biochemical features of Cdc13 allow it to perform its biological roles, we introduced a set of eight alanine mutations into the DBD across the binding interface (16, 20), with an emphasis on the aromatic residues that play key roles in affinity and specificity, and measured the impact of these changes on binding characteristics (Fig. 1). The change in binding affinity to the minimal Tel11 substrate (GTGTGGGTGTG) exhibited by these mutant proteins was measured at the physiological salt conditions identified previously using an EMSA binding assay () (12). The DBD constructs exhibited a range of binding affinities, from slightly tighter than the very tight WT apparent Kd of 2.1 pM to a reduced value of 71 pM (Fig. 1 and ). These defects in binding cannot be attributed to a change in protein structure or stability. Circular dichroic and NMR analysis suggest no alterations in secondary or tertiary structure (). Furthermore, most of the mutations did not significantly alter the melting temperature of the protein, and the observed minor changes show no correlation with biochemical activity, presumably because they are all well above the temperature at which the binding and in vivo studies were conducted (). Thus, the impact on the binding affinity to the Tel11 substrate exhibited by these mutant proteins spanned almost 35-fold, creating a set of proteins exhibiting a wide range, or rheostat, of binding affinities.

Large Defects in Binding Affinity Correlate with Substantial Impacts on in Vivo Viability.

This range of binding affinities allowed us to ask whether the unusually tight affinity exhibited by Cdc13 was required and, indeed, what level of DNA binding was necessary, for function in vivo. To do so, the mutations described above were examined for their effects in vivo by integrating each mutation into the genome of a diploid strain of yeast in place of one copy of the WT CDC13 gene. This panel of diploid strains was used to generate cdc13-DBD− haploid strains, which revealed a gradient of viability (Fig. 2). Changes in viability were not explained by changes in protein levels ().

Fig. 2.

Mutations in the Cdc13 DBD interface exhibit a gradient of viability in vivo. (A) Viability of yeast strains bearing the indicated cdc13-DBD− mutations was assessed by monitoring their ability to form visible colonies (or microcolonies) following sporulation and tetrad dissection of cdc13-DBD−/CDC13 diploid strains, with or without exo1-∆/EXO1 or rad9-∆/RAD9, to generate haploid strains with the specified genotypes. The resulting haploid strains were grown at 30° for 48 h unless otherwise indicated. Photographs were taken with a Zeiss Axioskop 50 microscope with a Nikon Digital Sight DS-5M camera, as described previously (23). Multiple isolates of each genotype were examined, and representative examples are shown. (B) Two isolates each of the haploid strains of the indicated genotypes were streaked onto rich medium and photographed after growth for 48 h at 30 °C to assess the extent of visible colony formation. Strains that were telomerase-proficient (TLC1) or telomerase-deficient [tlc1-∆, with a deletion of the telomerase RNA gene (50)] were generated by sporulation and tetrad dissection of isogenic CDC13/CDC13 tlc1-∆/TLC1, cdc13-Y561A/CDC13 tlc1-∆/TLC1, and cdc13-I578A/CDC13 tlc1-∆/TLC1 diploid strains.

Not unexpectedly, the cdc13-Y522A and cdc13-K622A mutant strains, which exhibited greatly reduced binding, were capable of only two to five cell divisions (Fig. 2), consistent with the 15- to 34-fold reduction in binding affinity for the minimal Tel11 substrate associated with these two mutations (Fig. 1). For both strains, this severe growth defect was partially suppressed by exo1-∆ and rad9-∆ mutations (Fig. 2); this recapitulates the behavior of previously characterized cdc13 mutations (21–23), arguing that defects in Cdc13 DNA binding behaved in a manner comparable to other loss-of-function mutations in CDC13. Notably, the growth of the cdc13-K622A mutant strain (with a 15-fold reduction in binding affinity) was reproducibly less impaired than the cdc13-Y522A strain (with a 34-fold reduction in in vitro binding), providing a strong correlation between the in vitro biochemical properties of these two mutations and their in vivo phenotypes.

Moderate Defects in Binding Affinity only Partially Correlate with in Vivo Viability.

Surprisingly, the correlation between in vitro and in vivo behavior did not extend to other mutations introduced into the DBD interface of Cdc13. For example, two mutations, cdc13-F539A and cdc13-Y626A, with only modest declines in affinity for the Tel11 substrate, nevertheless exhibited pronounced growth defects. Both these haploid mutant strains gave rise to barely visible colonies (Fig. 2 and ) which were accompanied by a high percentage of inviable individual cells for both strains, resulting in a long delay in forming visible colonies. Thus, despite having only a 3.5-fold effect on in vitro binding affinity, the cdc13-F539A and cdc13-Y626A mutant strains exhibited a severe degree of in vivo telomere dysfunction.

Mutations That Increase Cdc13-Binding Affinity Show Growth Defects in Strains Sensitized to Telomere Dysfunction.

Equally striking was the behavior of strains expressing mutant proteins that slightly increased the affinity for the Tel11 substrate relative to the affinity of the WT protein for Tel11 (Fig. 1 and ). In an otherwise WT yeast background, strains bearing mutations in Y556, I578, or Y561 exhibited a growth phenotype that was indistinguishable from that of a WT strain (Fig. 2 and ). However, when cdc13-Y561A or cdc13-I578A mutations were introduced into a strain background that is impaired for an additional aspect of telomere homeostasis (a telomerase deficiency), these mutant proteins were incapable of conferring the same level of function as the WT Cdc13 protein. Immediately following the loss of telomerase, the growth of a telomerase-defective strain is initially indistinguishable from that of a telomerase-proficient strain, although a decline in viability eventually becomes evident with continued propagation (24). In contrast, a newly generated telomerase-defective strain that also bore either a cdc13-Y561A or a cdc13-I578A mutation exhibited an immediate decline in viability (Fig. 2). Similarly, these same mutations also exhibited a pronounced synthetic growth defect when combined with a mutation in the Ku heterodimer; the cdc13-Y556A yku80-∆, cdc13-I578A yku80-∆, and cdc13-Y561A yku80-∆ double-mutant strains were close to inviable (). The synthetic lethalities due to these mutations in the DBD interface were not readily explained by their small increase in affinity for the Tel11 substrate.

Binding Specificity Is Profoundly Altered by Mutations in the DNA-Binding Interface.

The above results strongly suggested that affinity was not the only important biochemical feature required for Cdc13 function in vivo. We therefore asked whether an additional biochemical property, binding specificity for telomeric substrates, was altered by these mutations. We have previously assessed Cdc13 specificity by measuring binding affinities for oligonucleotides with substitutions for the “pool” of the three other bases at specific positions within the minimal Tel11 oligonucleotide (10). This approach revealed a “specificity profile” defined by the relative loss of affinity when the identity of a base in the ligand is altered. The larger the loss in affinity for the pool relative to the cognate ligand, the more specifically the cognate base is recognized. This strategy revealed that bases at positions G1, G3, and T4 within the Tel11 (GTGTGGGTGTG) substrate are the most specifically recognized by both the Cdc13 DBD and the full-length Cdc13 protein (10, 12). Substitutions at these three positions in the Tel11 sequence led to a significant loss of affinity (up to 87-fold) by the WT protein, whereas the change in affinity upon substitution at G9, a site which is less specifically recognized, was more modest (Table 1 and ).

Table 1.

Apparent Kd values for WT and mutant Cdc13 DBD proteins to Tel11 variants

Protein	Tel 11, pM	H1*, pM	H3*, pM	V4*, pM	H9*, pM
WT	2.1 ± 0.2	31 ± 5	85 ± 20	180 ± 40	9 ± 1
Y556A	0.6 ± 0.1	300 ± 30	310 ± 50	570 ± 200	43 ± 4
I578A	1.0 ± 0.2	420 ± 20	130 ± 20	280 ± 70	20 ± 2
Y561A	1.6 ± 0.2	15 ± 30	430 ± 80	380 ± 30	16 ± 2
Y565A	4.7 ± 0.8	330 ± 50	260 ± 20	370 ± 70	34 ± 5
Y626A	7.2 ± 0.9	1,230 ± 60	2,500 ± 600	290 ± 70	180 ± 10
F539A	7.4 ± 0.2	1,200 ± 100	2,600 ± 400	1,300 ± 100	180 ± 10
K622A	31 ± 3	2,400 ± 400	1,200 ± 200	4,500 ± 1,000	210 ± 60
Y522A	71 ± 16	430 ± 50	770 ± 200	6,000 ± 1,000	1,200 ± 300

H refers to an equimolar pool of A, C, and T; V refers an equimolar pool of G, C, and A.

To determine how specificity is impacted by mutations across the DBD interface, binding to these pools of oligonucleotides was performed with all mutant proteins (Fig. 3, Table 1, and ). A wide range of effects was observed when the pool of bases was substituted at positions in the Tel 11 oligo, with the reductions in affinity ranging from 4.5-fold to nearly 3,000-fold.

Fig. 3.

Creation of a rheostat of binding affinities in Cdc13. (A) Skyscraper plot of Kds for each alanine mutant of Cdc13 DBD mutant protein with various substituted oligonucleotides. (B) Specificity data for select mutant proteins scaled to the loss of affinity observed for that mutant to Tel11. Data used to construct this plot are in .

A standard double-mutant thermodynamic cycle can be used to assess the effects of independently mutating the protein or the nucleic acid substrate and then combining these two different alterations to the protein/nucleic acid interface (25–27). If these are independent changes, the effects of implementing them simultaneously on binding free energy will simply be additive, with the net observed Kds being the product of the Kds for the individual changes. However, if the combination has a nonadditive effect on affinity, it suggests that the two alterations are thermodynamically coupled in some way. A net effect that is less than simply additive could be due to the sites being physically proximal, so that the loss represents both sides of a direct amino acid/base contact and removing either side of the interaction is sufficient to abrogate it. For this reason, strong couplings are most commonly observed for physically proximal residues (25–27). Moreover, it could also suggest that a mechanism of accommodation is in place whereby the loss of a favorable interaction is compensated by the gain of a new favorable interaction, as seen previously in other ssDNA/protein complexes (28, 29). Conversely, a net effect that is greater than the sum of the free energy changes of the individual alterations could suggest a loss of cooperativity or a structural change at the interface. This classic analysis allows us to identify the amino acids that perform unexpected roles in determining binding specificity.

Simple thermodynamic additivity explained the binding affinities observed for several double-mutant pairs. For example, the weakest binding mutant of Cdc13 DBD, Y522A, which has been previously designated as a hotspot for binding affinity, showed the most substantial decline in binding to the WT Tel11 ligand (34-fold, to 71 pM). Similarly, substitutions at the most specifically recognized site on the oligonucleotide, T4, resulted in an 86-fold decline in affinity, to 180 pM, for the WT protein. If these effects were simply additive, when assessing binding of the mutant protein (Y522A) to the modified ligand the reduction in affinity would be predicted to be 2,958-fold, which is quite similar to the observed value of 2,800-fold (). The fact that these sites behave independently is consistent with their ∼15-Å separation in the structure (30). Y522, however, is physically proximal to G1. Here, the impact on binding in the doubly substituted Y522A/H1 complex was less than additive, with the observed net affinity down 200-fold relative to WT, whereas the additive effect would be ∼500-fold. This deviation from a simply additive result supports the prediction from the structure that Y522 specifically recognizes G1.

To visually identify protein/nucleic acid pairs whose combined alteration deviates from thermodynamic additivity, we divided the affinities for the binding of the doubly substituted pairs by the binding affinity of each mutant protein for Tel11 (Fig. 3 and ). Thermodynamic additivity would predict that, with this normalization, the mutant proteins would show the same specificity profiles as observed for WT Cdc13. This scaled specificity profile indeed revealed that, in mutants for which the effects were close to additive, such as Y522A, the specificity profile mirrored that of WT. Exceptions included sites of direct contact where the impact on binding was smaller than expected, as discussed above for the Y522A/H1 pair.

Several mutant Cdc13 proteins unexpectedly exhibited a deviation from additivity in which the pairs led to a greater loss of affinity than predicted by additivity. This was particularly evident for Y626A and F539A and to a lesser extent for Y561A (Fig. 3 and ). A case in point was the interaction of F539A with H1, where the combined reduction in affinity (600-fold) was about 11-fold greater than the product of the F539 vs. Tel 11 (3.5-fold) and WT vs. H1 (15-fold) differences between WT binding with Tel 11. This was also the case for the interaction between F539A and position H3 in the DNA: F539A exhibited a 1,300-fold loss in binding affinity at this site even though F539A exhibited only a 3.5-fold reduction in Tel11 binding. Again, the net impact of the combination of amino acid change and oligonucleotide substitution was highly nonadditive. This binding profile for the F539A mutation argues that the Cdc13-F539A protein was more specific for the Tel11 sequence, as substitution of the oligonucleotide base led to greater losses in binding than observed for the WT protein. A key observation is that this enhanced specificity was not due to tighter binding of the mutant proteins to Tel11; rather, it arose from a decreased tolerance for the substitutions in the oligonucleotide. Notably, these effects were not manifest only at the base closest to the site of alanine mutation in the structure (Fig. 1), suggesting that long-range effects across the DBD interface dictate binding specificity. F539A illustrates this nicely: Although this substitution had a large, nonadditive impact on H1 and H3 binding, it is poised between T4 and G5 in the structure of the complex (30).

Binding Specificity in Conjunction with Affinity Predicts in Vivo Phenotypes More Accurately.

We noted above that in vivo phenotypes correlated roughly with severe losses of binding affinity but that more moderate changes in binding affinity did not fully explain the phenotypes. The reduced tolerance by F539A and Y626A for deviations from the Tel11 sequence (i.e., increased specificity) provides a biological explanation for their in vivo phenotypes, which were significantly functionally impaired, particularly when contrasted with the similarly affinity-impaired Y565A. The severe growth defects associated with the cdc13-F539A and cdc13-Y626A strains, as well as the more subtle growth defects in the cdc13-I578A, cdc13-Y561A, and cdc13-Y556A yeast strains (with mutations that conferred increased affinity but reduced specificity) demonstrate that binding specificity contributes substantially to Cdc13 function.

Discussion

In this study, we have performed a systematic analysis of the ssDNA-binding surface of Cdc13 by generating a panel of alanine mutations that span the interface and then probing the impact of these mutations on binding affinity and specificity. This detailed biochemical analysis was combined with an in vivo phenotypic read-out that was sensitive enough to detect even minor differences in function, revealing a gradient, or rheostat, of functionality. As expected, strains expressing mutant proteins with a reduction in binding affinity of more than 15-fold were inviable, demonstrating that high-affinity DNA binding is an essential function of the yeast t-RPA complex which contains the Cdc13 protein. Surprisingly, this systematic analysis identified a second category of mutations that did not confer substantial changes in ssDNA-binding affinity but altered the ssDNA-binding specificity of the Cdc13-binding interface, such that the surface was less tolerant of changes in the ssDNA. This second category of cdc13 mutations also had a substantial impact on viability, thereby revealing that sequence tolerance is as important as binding affinity for biological function in vivo.

Typically, mutating a contacting amino acid increases sequence tolerance by removing the H-bond donors and acceptors and steric interactions that enforce specific recognition. In contrast, mutations in a subset of residues of the Cdc13-binding interface confer a decrease in sequence tolerance. Notably, mutating these amino acids impacts recognition of bases 10 Å away (Fig. 1), indicating that long-range effects across this interface contribute to specificity. Furthermore, removal of these aromatics from the interface does not make binding more promiscuous, suggesting that these side chains are not driving local specificity and instead are accommodating sequence diversity. We therefore propose a model in which DNA binding by Cdc13 employs a highly cooperative interface with sequence diversity accommodated through plastic binding modes. This argues that ssDNA binding employs localized contacts between a subset of amino acids and adjacent bases that are important for binding affinity as well as long-range effects across the interface that are critical for sequence tolerance.

Analysis of the biochemical data in the context of the Cdc13 DBD/Tel11 structure points to three distinct functional parts of the interface (30). The first region is the segment of the OB-fold barrel that interacts with the 5′ end identified in previous mutagenesis studies as driving both affinity and specificity of interaction. This region includes Y522 and K622, the residues whose substitution has the largest impact on affinity without significant changes in specificity. The second is the long β2–3 loop (highlighted in blue in Fig. 1), encompassing mutations spanning residues Y556A to Y565A, that interacts with the 3′ end of the ligand. Mutations in this loop have more moderate impacts on affinity and specificity, consistent with a “Velcro-like function,” that is, a sticky surface suited to binding any sequence. The final structural region bridges these two, spanning the middle part of the barrel. Here substitution of two key aromatic residues, F539A and Y626A, results in a modest loss in affinity but a significant increase in specificity characterized by the dramatic loss of tolerance of substitutions at the rather distant sites of G1 and G3. Thus, this middle region appears to control the plasticity of the recognition so that Cdc13’s ability to accommodate sequence alterations is impaired upon loss of these aromatic residues. The behavior is reminiscent of another sequence-tolerant telomere end-binding protein, Pot1pC of Schizosaccharomyces pombe, in which sequence tolerance is implemented through new binding modes that thermodynamically compensate for base substitutions through alternate stacking interactions and new H-bonding networks (28). Our data suggest that the loss of key aromatic residues in this middle region impairs the ability of the protein to tolerate alternative sequences, perhaps due to the ability of the aromatic amino acids to stack on the exposed bases of ssDNA and affect plasticity (31).

In vivo, the phenotypes displayed by strains bearing mutations in the DBD interface showed numerous similarities to previously described mutations in CDC13 that confer viability defects, arguing that this set of cdc13-DBD− mutations is impacting the primary Cdc13 function. Severely impaired cdc13-DBD− strains displayed a DNA damage response and impaired cell-cycle progression () comparable to that of previously characterized cdc13-defective strains (32). This panel of cdc13-DBD− mutant strains also exhibited a profile of genetic interactions in response to rad9-∆ and exo1-∆ mutations (Fig. 2) that recapitulated the behavior of cdc13 temperature-sensitive (cdc13-ts) strains (23, 32). However, there was one notable difference. In cdc13-ts strains grown at nonpermissive temperatures, there is a marked increase in the extent of exposed telomeric G-strand ssDNA (33, 34). This observation, combined with the enhanced DNA damage response observed in cdc13-impaired strains, led to the premise that telomeric ssDNA creates a specific signal that elicits a cell-cycle checkpoint (32, 33). In contrast, none of the cdc13-DBD mutant strains—not even those that were severely impaired—exhibited any detectable increase in the extent of G-strand ssDNA at chromosome ends (). This suggests that the primary DNA lesion eliciting a checkpoint response in cdc13-impaired cells may not be ssDNA but in fact may be some other intermediate that arises during DNA replication stress.

The behavior of yeast strains expressing three mutations highlights the in vivo importance of rigorously tuning specificity. F539A, Y565A, and Y626A all exhibit similar reductions in affinity (2.2- to 3.5-fold) (Fig. 1) but vary markedly with regard to specificity (Fig. 3). Y565A is modestly more specific than WT, while F539A and Y626A are significantly more specific. This in vitro gradient of specificity generates a comparable in vivo gradient, as a strain expressing the mutant Cdc13-Y565A protein is slightly less functional than WT (), whereas cdc13-F539A and cdc13-Y626A are severely impaired (Fig. 2 and ). This surprising result shows that a gain in specificity can actually be deleterious to function in vivo. It also underscores that DNA-binding specificity is not a binary (specific vs. nonspecific) trait; rather there is a continuum of specificity that is critical to the biological functioning of many DNA-binding proteins. Specifically, our results indicate that the recognition of ssDNA by Cdc13 relies on a finely tuned balance of both affinity and specificity to ensure that the t-RPA complex can readily localize to a limited region of the genome and still accommodate the sequence heterogeneity present at yeast telomeres.

Specificity in nucleic acid recognition by proteins has been studied from both biochemical and structural perspectives (reviewed in ref. 35), which have suggested that the nucleic acid-binding interface is malleable. For example, single-stranded recognition interfaces can be remodeled to match different substrates to achieve specific recognition for both DNA and RNA, as exemplified by the Oxytricha nova telomere end-binding protein (36), the S. pombe Pot1 protein (28, 37), and, in RNA recognition, the PUF protein (38) and the MS2 coat protein (39). This malleability could be dynamic in origin (40, 41). Lacking in all these prior studies, however, has been a demonstrated link between the requirement for recognition malleability and function.

The observation that mutations in the DNA-binding interface of Cdc13 render the protein more specific and less functional was unexpected. While it is common to observe loss of function upon loss of a biochemical activity, the enhancement of specificity leading to a substantial reduction in biological function has not been reported previously in nucleic acid recognition, to the best of our knowledge. As a systematic evaluation of the binding specificity of mutant proteins is not commonly undertaken, this disruption of multiple biochemical behaviors may be a broader phenomenon than previously appreciated. A case in point is the human CST complex, a heterotrimer with a domain organization very similar to that of the yeast t-RPA complex (42–44). Unlike t-RPA, the CST complex is not a telomere-dedicated protein; although it displays a preference for G-rich sequences, the arrangement of guanosine nucleotides needed for high-affinity binding does not correspond to the repeat characteristic of telomeres (45–48). This allows the complex to function as a replication accessory factor genome-wide as well as facilitating proper maintenance of G-rich sequence at telomeres. Although the complete DNA-binding interface of CST has not yet been identified, we suggest that the results reported here for the yeast Cdc13 protein may extrapolate to other modestly specific ssDNA-binding complexes such as CST. Moreover, perhaps ssDNA-binding proteins, such as RPA and SSB, which are largely nonspecific, achieve nonspecific binding through a similar mechanism, as suggested by dynamic analysis of the RPA-binding domain (49). As the alteration in the specificity of Cdc13 was discovered only through the comprehensive mutagenesis of the protein surface that contacts DNA, it suggests that the systematic mutation and characterization of an entire interaction surface is essential to understand the full complexity underlying nucleic acid binding and in vivo function.

Methods

Protein expression, purification, binding and specificity studies, and in vivo analysis are described in detail in .