Phosphopeptide Binding Specificities of BRCA1 COOH-terminal (BRCT) Domains*

Protein phosphorylation by protein kinases may generate docking sites for other proteins. It thus allows the assembly of signaling complexes in response to kinase activation. Several protein domains that bind phosphoserine or phosphothreonine residues have been identified, including the 14-3-3, PIN1, FHA, KIX, WD-40 domain, and polo box (Yaffe, M. B., and Elia, A. E. (2001) Curr. Opin. Cell Biol. 13, 131-138; Elia, A. E., Cantley, L. C., and Yaffe, M. B. (2003) Science 299, 1228-1231). The BRCA1 COOH-terminal (BRCT) domains are protein modules found in many proteins that regulate DNA damage responses (Koonin, E. V., Altschul, S. F., and Bork, P. (1996) Nat. Genet. 13, 266-268). Whether BRCT domains can mediate phosphorylation-dependent interactions has not been systematically investigated. We report here that the BRCT domains also recognize phosphopeptides. Oriented peptide library analysis indicated that the BRCT domains from BRCA1, MDC1, BARD1, and DNA Ligase IV preferred distinct phosphoserine-containing peptides. In addition, the interaction between BRCA1 and the BRCT binding motif of BACH1 was required for BACH1 checkpoint activity. Furthermore, BRCT domains of the yeast DNA repair protein Rad9 could bind phosphopeptides, suggesting that the BRCT domains represent a class of ancient phosphopeptide-binding modules. Potential targets of BRCT domains were identified through data base search. Structural analysis of BRCA1 BRCT repeats also predicted conserved residues that may form the phosphopeptide-binding pocket. Thus, the BRCT repeats are a new family of phosphopeptide-binding domains in DNA damage responses.

Originally identified as a sequence motif homologous to the BRCA1 COOH-terminal region, the BRCT 1 domain is a protein domain approximately 90 amino acids in length (3)(4)(5)(6). It is an ancient protein module that can be found in single cell eukaryotes (7). In the human genome more than 30 BRCTcontaining proteins have been documented. Interestingly, most of the BRCT domains associate with proteins involved in DNA repair and cell cycle checkpoint response (5,6). The exact function of BRCT domains is not yet clear; they have been postulated to mediate protein-protein interactions (7,8). Heterodimerization between single BRCT domains (e.g. XRCC1 and DNA Ligase III) has been reported (9,10). In addition, BRCT domains of 53BP1 can interact with P53, while BRCT domains of BRCA1 bind DEAH family helicase BACH1 and CtBP interacting protein CTIP (11)(12)(13). Interestingly, it was shown that the BRCT domains might also bind double stranded DNA breaks (14).
In addition to heterodimerization of BRCT domains, emerging evidence suggests that BRCT domains may mediate phosphorylation dependent interactions. For example, association of the sixth BRCT domain of TOPBP1 and E2F1 was shown to be regulated by phosphorylation (15). Interactions between BRCA1 and BACH1 depend on BACH1 phosphorylation as well (16). However, whether other BRCT domains can directly binding to phosphopeptide and what the specificities of BRCT domains are remain to be determined.
We were particularly intrigued by the possibility of BRCT domains mediating phosphorylation-dependent interactions, because protein phosphorylation by protein kinases is known to trigger the assembly of phosphorylation-dependent signal complexes critical for cell growth and survival (1,17). Many protein domains or modules have evolved to mediate such activities.
The best examples are the Src homology domain 2 and phosphotyrosine-binding domains that recognize phosphorylated sites by protein-tyrosine kinases in a sequence-specific manner (17,18). Recently, several protein domains that bind phosphoserine or phosphothreonine residues have been identified, including the 14-3-3, FHA, KIX, PIN1, WD-40, and polo box (1,2). Notably, each of these domains recognizes a subset of substrates phosphorylated by different protein kinases. For instance, 14-3-3 recognizes sites that are phosphorylated by Argdirected kinase such as Akt/PKB, while FHA domains bind sites that are substrates of phosphatidylinositol 3-kinase family (1). Given the large number of protein kinases in the human genome, it has been postulated that novel families of phosphopeptide-binding domains that transmit signals from different protein kinases may exist. If BRCT domains can indeed bind phosphopeptides directly, it will provide new insight about how signaling complexes are regulated by DNA damage. To answer such questions, we carried out experiments to study the specificities of BRCT domains using oriented peptide libraries.
Oriented Peptide Library Analysis-Two phosphoserine-containing peptide libraries (X3: ISRSTpSXXXNK and X6: KAXXXpSXXXAK) were synthesized as described previously (19), where X is any amino acids except Cys. For library analysis, 1 mg of the peptide library was incubated with 100 -300 g of GST BRCT-agarose beads. Then the beads were washed five times with 2 ml of phosphate-buffered saline. The bound peptides were eluted with 30% acetic acids, lyophilized, and sequenced on an ABI 477 peptide sequencer. Relative selectivities of amino acids at each degenerate position were calculated as described previously (19).
SPOT Peptide Array Synthesis and Screen-The BACH1 peptide array was synthesized on cellulose membrane using the ASP222 SPOT robot (20). The parental peptide SRSTpSPTFNK was scanned with 19 of the 20 amino acids (except Cys), phosphoserine, and phosphothreonine. The array membrane was first blocked with 3% bovine serum albumin in TBST (0.1 M Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20) for 2 h at room temperature. GST-BRCT of BRCA1 (3 g) was labeled with anti-GST-horseradish peroxidase (Amersham Biosciences) (0.75 g) for 30 min at room temperature. The fusion proteins were then added to the array membrane at a final concentration of 2 g/ml for 30 min. The array membrane was subsequently washed with TBST for 10 min three times. GST-BRCT bound peptide spots were visualized by ECL.
Affinity Measurements Using Fluorescence Polarization-A FITClabeled phosphopeptide with the sequence VDDpSYVFNK was first synthesized. For controls, an aliquot of this peptide was dephosphorylated with alkaline phosphatase. The phosphopeptide or dephosphorylated peptide was incubated with different concentration of BRCA1 GST-BRCT fusion proteins in a 96-well plate. Fluorescence polarization was measured on TECAN Polarion.

RESULTS
The BRCA1 BRCT Domains Recognize Phosphopeptides with the Phosphoserine-Aromatic-Hydrophobic-Phe Motif-One of the known in vivo targets of BRCA1 is BACH1, a member of the DEAH helicase family (13). In the course of mapping BRCA1 and BACH1 interactions, we found that tandem but not individual BRCA1 BRCT domains directly bound BACH1 in a phosphorylation-dependent manner (16). Further analysis of BRCA1 and BACH1 interaction demonstrated that BRCA1 BRCT domains specifically recognize phosphorylated Ser 990 (ISRSTS 990 PTFNK) of BACH1. These data suggested that the BRCT domain may be a phosphopeptide-binding module.
To understand the sequence binding preference of BRCT domains, we first examined the specificity of BRCA1 BRCT domain using an oriented peptide library (19). The phosphoserine containing peptide library X3 (ISRSTpSXXXNK) was synthesized, based on the sequence surrounding residue Ser 990 of human BACH1. Using this phospho-library, we found that BRCA1 BRCT domains preferred aromatic amino acids at the Pϩ1 position and aromatic/hydrophobic residues at the Pϩ2 position (Fig. 1). Notably, at the Pϩ3 position COOH-terminal to the phosphoserine, Phe is strongly selected, indicating that Phe at this position is critical for recognition by the BRCA1 BRCT domains. Thus, BRCA1 BRCT domains prefer a phosphoserine-aromatic-hydrophobic-Phe motif.
To further examine the specificity of BRCA1 BRCT domain at residues NH 2 -terminal to the phosphoserine, the phospho-library X6 with the sequence KAXXXpSXXXAK was used. As shown in Table I 2 terminus, suggesting that residues COOH-terminal to phosphoserine are important for binding to BRCA1 BRCT domains. A BRCA1 BRCT binding peptide (VDDpSYVFNK) was then synthesized, and its affinity to BRCA1 BRCT domains was measured. Indeed, this peptide bound with an affinity of 162 nM to the BRCA1 BRCT domains. In addition, such interaction was dependent on phosphorylation, as the unphosphorylated control peptide failed to bind ( Fig. 2A).
To further probe the binding motif for BRCA1 BRCT domains, we generated an array of phosphopeptides based on the FIG. 1. The BRCT repeats of BRCA1 recognize specific phosphoserine-containing peptides. The binding specificity of BRCA1 BRCT domains was determined using an oriented peptide library ISRSTpSXXXNK. Relative selections of individual amino acids at position COOH-terminal to phosphoserine (Pϩ1, ϩ2, and ϩ3) were shown here. A value smaller than 1 indicates that the amino acid is not preferred. Trp was not included due to peptide sequencing problems. BACH1 peptide using the SPOT peptide synthesizer. Consistent with our finding that Phe at Pϩ3 is crucial for binding, substitution with amino acids other than Phe at Pϩ3 abolished BRCA1 BRCT interaction (Fig. 2B). Similarly, only ␤-branched amino acids (Thr, Val, and Ile) were tolerated at the Pϩ2 position. This result recapitulated our oriented peptide library data and again highlighted the importance of the Pϩ3 position for recognition by the BRCA1 BRCT domain. The oriented peptide library and SPOT array data collectively demonstrate that the BRCT domains from BRCA1 recognize specific phosphorylated sequences with the consensus motif phosphoserinearomatic-hydrophobic-Phe motif.
The BRCT-binding Motif of BACH1 Is Important for BRCA1-dependent G 2 /M Checkpoint Control-BRCA1/BACH1 interaction is important for DNA damage-induced checkpoint control during G 2 to M cell cycle transition (16). To test whether the BRCA1-binding motif of BACH1 identified by in vitro studies is involved in this process, we used a ␥-ray-induced G 2 /M checkpoint assay.
First, we constructed Thr 989 -to-Ala (T989A), Ser 990 -to-Ala (S990A), or Phe 993 -to-Ala mutants of BACH1. BACH1 siRNAresistant forms of wild-type and mutant BACH1 were then generated by introducing silent mutations in the siRNA target region. All these BACH1 proteins were expressed in HeLa cells. However, only wild-type BACH1 and the T989A mutant, but not the S990A or F993A mutant, were associated with BRCA1 (Fig. 3A). This result indicates that the BRCA1 binding-motif of BACH1, especially Pϩ3 Phe, is necessary for BRCA1/BACH1 interaction in vivo. To explore the in vivo functional link between the BRCT-binding motif of BACH1 and BRCA1 BRCT domains, HeLa cells were treated with BACH1 siRNA and then transfected with siRNA-resistant BACH1 or BACH1 mutants. BACH1 siRNA repressed the expression of endogenous BACH1 but not exogenous siRNA-resistant BACH1 (Fig. 3B). Cells treated with BACH1 siRNA showed G 2 /M checkpoint defect after DNA damage (Fig. 3C). While expression of wild-type BACH1 and the T989A mutant rescued the checkpoint defect, the S990A and F993A mutants had no effect (Fig. 3, D and E). These data further confirm Pϩ3 Phe as a key residue for BRCA1 BRCT and BACH1 interaction and indicate that the BRCA1 binding-motif of BACH1 is important for the function of BRCA1 BRCT domain in vivo.
The Specificities of BRCT Domains of MDC1, BARD1, and Ligase IV-To investigate whether other BRCT domains can also bind phosphopeptides, we studied the specificities of several BRCT domains. Interestingly, tandem BRCT domains from MDC1, BARD1, and Ligase IV specifically bound phosphopeptides in the X3 library (Table I).
MDC1 is a recently identified mediator of DNA damage responses (21)(22)(23)(24)(25)(26). MDC1 contains two BRCT repeats at its carboxyl terminus. Using the X3 library, we found that MDC1 BRCT domains selected a distinct set of phosphopeptides compared with BRCA1. Instead of Phe, Tyr is the most selected amino acid by MDC1 at the Pϩ3 position (Table I) BARD1 is a cancer susceptible protein that is mutated in families of breast cancer patients (27). BARD1 has an NH 2terminal ring finger domain which heterodimerizes with BRCA1 and two COOH-terminal BRCT domains with unknown function (4). Oriented peptide library analyses indicated  that the tandem BRCT domains of BARD1 specifically recognize Asp/Glu residues at the Pϩ1, ϩ2, and ϩ3 positions ( Table  I) DNA Ligase IV functions to join single-strand breaks in doublestranded DNA (28,29). It plays a major role in V(D)J recombination and non-homologous end joining. At the COOH terminus of Ligase IV are two BRCT domains. Different from MDC1 and BARD1, Ligase IV BRCT repeats preferred to bind phosphopeptides with Tyr and Ile, respectively, at the Pϩ1 and Pϩ3 positions (Table I). These results demonstrated that the BRCT domains are protein modules that recognize specific phosphopeptides.
S. cerevisiae RAD9 BRCT Domains Bind Phosphopeptides-The BRCT domain is an ancient structural fold that can be traced back to bacteria. In bacteria and fungi, BRCT domains are also frequently found in proteins that regulate DNA damage responses. We therefore were interested in determining if BRCT domains from single cell eukaryotes could bind phosphopeptides as well. One of the proteins we studied is budding yeast RAD9, which is essential for cell cycle checkpoint control (30). As shown in Table I, the tandem BRCT domains of RAD9 preferred Tyr at the Pϩ1 position, and Ile at the Pϩ2 and Pϩ3 positions. Thus, the RAD9 BRCT repeats recognize phosphopeptides with the motif pS-Y-I-I. This result suggests that BRCT domains have resistant mutants rescued the G 2 /M checkpoint defect generated by BACH1 siRNA but not the S990A or F993A mutant. E, the fraction of M phase cells from C and D is expressed as a percentage of that measured in non-irradiated control cells.  Prediction of Potential in Vivo Targets of BRCT Domains-To identify potential targets of BRCT domains, we performed pattern searches in Swiss-Prot using the consensus motifs from our library and SPOT array analyses (Table II). Some of the interesting predicted targets for BRCA1 include BRCA1 itself, kinesin-like protein KIF1B, mismatch repair protein Msh3, Rb-binding protein CTIP, co-repressor NCoA-2 and NCoA-3/SRC-3. For MDC1, histone deacetylases HDAC8, and HDAC10, GATA family transcription factors could be potential targets. Interestingly, RNA polymerase II contains repeats of the MDC1-binding motif. A search of potential targets of RAD9 in budding yeast also found a short list of proteins that regulate DNA repair and cell cycle. Notably these include MEC1, RAD1, and RFC p140, which have been shown to interact with RAD9 genetically and biochemically.

DISCUSSION
Tandem Versus Single BRCT Domains-Our results have shown that tandem BRCT repeats can specifically recognize phosphoserine containing peptides, suggesting that the BRCT domain represents a new class of modules that mediate phosphorylation dependent protein-protein interactions. Most BRCT domain-containing proteins are known to mediate DNA damage responses (5,6). These proteins are often recruited to the sites of DNA damage, coinciding with the activation of protein kinases such as ATM, ATR, Chk1, and Chk2 (31)(32)(33). The identification of BRCT domains as phosphopeptide-interacting motifs suggest that BRCT domains may facilitate the assembly of signaling complexes that sense the activation of damage and checkpoint kinases. In fact, numerous residues in the BRCA1 BRCT repeats are mutated in breast cancer cells (34,35). Some of these mutations may disrupt the ability of BRCT domains to bind phosphorylated cellular target and therefore contribute to disease phenotypes.
There is one notable difference between BRCT and other phosphopeptide-binding domains such as Src homology domain 2 and FHA domains. Among the BRCT domains we examined so far, only tandem but not single BRCT domains bound a sufficient amount of phosphopeptides from the peptide library. It is possible that inter-BRCT interactions is required for phosphopeptide binding, or tandem BRCTs function as a unit to bind phosphopeptides. In fact, structural studies have revealed that tandem BRCTs may have evolved from a common ancestor, because the linker structures between the two BRCT domains are also conserved (36). In further support of this idea, single point mutations in either of the two repeats in BRCA1 can abolish its interaction with BACH1 (13). Our data, however, do not rule out that single BRCT domains can also bind phosphopeptides. Interestingly, single BRCT domains (e.g. XRCC1 and Ligase III) have been shown to form heterodimers. One intriguing possibility may exist that such heterodimers may function as a phosphopeptide-binding unit.
Structural Insight into BRCT Phosphopeptide Recognition-The structures of several BRCT domains have been determined (8, 13, 36 -38). These studies indicate that the BRCT domain consists of four-stranded parallel ␤ sheets bundled by three ␣ helices. The two BRCT repeats of BRCA1 connected by the linker region are arranged in tandem (36). It is possible that similar to 14-3-3 and FHA, the phosphoserine-binding pocket of BRCT repeats may be formed by conserved Arg or Lys residues (39,40). Secondary structure comparisons of BRCA1 BRCT repeats from several different species suggested that Lys 1667 , Arg 1670 , Lys 1671 , Lys 1724 , Arg 1726 , Lys 1727 , Arg 1737 , Lys 1750 , Arg 1751 , Arg 1753 , and Lys 1759 on human BRCA1 are highly conserved. In addition, positive charged residues are found at similar positions on BRCT domains of MDC1, BARD1, and DNA Ligase IV. It is interesting to note that these Arg/Lys residues are mainly located at two regions of BRCA1 BRCT domains. One region is on the surface of BRCT domain structure formed by Lys 1667 , Arg 1670 , and Lys 1671 located on the helix ␣1A, and Arg 1726 and Lys 1727 located on the loop between ␣3A and ␣4A. The second region is near the connecting linker of the tandem BRCTs. These regions may form the phosphoserine-binding pocket of BRCA1 BRCT repeats. In support of this model, mutation of several residues in the second region of the BRCA1 BRCT repeats significantly decreased BACH1 binding in vitro, suggesting that the phosphopeptide-binding pocket may reside in this region as well (36).