Report

BRCT Repeats As Phosphopeptide-Binding Modules Involved in Protein Targeting

See allHide authors and affiliations

Science  24 Oct 2003:
Vol. 302, Issue 5645, pp. 636-639
DOI: 10.1126/science.1088877

Abstract

We used a proteomic approach to identify phosphopeptide-binding modules mediating signal transduction events in the DNA damage response pathway. Using a library of partially degenerate phosphopeptides, we identified tandem BRCT (BRCA1 carboxyl-terminal) domains in PTIP (Pax transactivation domain-interacting protein) and in BRCA1 as phosphoserine- or phosphothreonine-specific binding modules that recognize substrates phosphorylated by the kinases ATM (ataxia telangiectasia–mutated) and ATR(ataxia telangiectasia– and RAD3-related) in response to γ-irradiation. PTIP tandem BRCT domains are responsible for phosphorylation-dependent protein localization into 53BP1- and phospho-H2AX (γ-H2AX)–containing nuclear foci, a marker of DNA damage. These findings provide a molecular basis for BRCT domain function in the DNA damage response and may help to explain why the BRCA1 BRCT domain mutation Met1775 → Arg, which fails to bind phosphopeptides, predisposes women to breast and ovarian cancer.

Signal transduction by protein kinases in eukaryotes results in the directed assembly of multiprotein complexes at specific locations within the cell (1). This process is particularly evident after DNA damage, where activation of DNA damage response kinases results in the formation of protein-protein complexes at discrete foci within the nucleus (2). In many cases, protein kinases directly control the formation of these multiprotein complexes by generating specific phosphorylated-motif sequences; modular binding domains then recognize these short phospho-motifs to mediate protein-protein interactions (3, 4).

We used a proteomic screening approach (5) to identify novel modular pSer- or pThr-binding domains involved in the DNA damage response. In cells exposed to γ-irradiation, the kinases ATM and ATR phosphorylate transcription factors, DNA repair proteins, protein kinases, and scaffolds on Ser-Gln and Thr-Gln motifs (6). We therefore constructed an oriented phosphopeptide library biased to resemble the motif generated by ATM and ATR (7, 8) (Fig. 1A). This library and its nonphosphorylated counterpart were immobilized and screened against ∼96,000 in vitro translated polypeptides (960 pools each containing ∼100 transcripts) (Fig. 1A).

Fig. 1.

Identification of pSer- or pThr-binding domains with an ATM/ATR-motif library versus an expression library screen. (A) An oriented (pSer/pThr) phosphopeptide library, biased toward the phosphorylation motifs of ATM and ATR, was immobilized on Streptavidin beads. The library [pSQ = biotin-ZGZGGAXXXB(pS/pT)QJXXXAKKK] and its nonphosphorylated counterpart (SQ) were screened against in vitro translated, [35S]Met-labeled proteins. Notation code: (pS/pT), 50% pSer and 50% pThr; Z, aminohexanoic acid; B, a biased mixture of the amino acids Ala (A), Ile, Lys (K), Met, Asn, Pro, Ser, Thr, and Val; J, a biased mixture of 25% Glu and 75% X, where X denotes all amino acids except Cys, His, Lys, and Arg (21); G, Gly; Q, Gln. In each panel, the first and second lanes show binding of proteins within the pool to the phosphorylated (pSQ) and nonphosphorylated (SQ) libraries, respectively. Identification of PTIP, denoted by arrow and asterisk, occurred through progressive subdivision of the EE11 pool to a single clone. The uppermost band is a fusion artifact of PTIP with vector sequences resulting in translation initiation at an upstream start codon in the vector. (B) Deletion mapping of the phospho-binding domain of PTIP. Truncations of PTIP were assayed for selective binding to the pSQ library as in (A). BRCT domain boundaries were determined from sequence alignments (22).

Pool EE11 contained the strongest phosphopeptide-binding clone, EE11-9, encoding the C-terminal 70% of human PTIP (Fig. 1B). PTIP, originally identified as a transcriptional regulatory protein, appears to also play a critical role in the DNA damage response pathway (911). Human PTIP contains four BRCT domains, known protein-protein interaction modules that are present in many DNA damage response and cell cycle checkpoint proteins (12). A construct containing only the tandem third and fourth BRCT domains displayed strong and specific binding to the (pSer/pThr)-Gln library (Fig. 1B). Constructs of PTIP lacking both of these domains either failed to bind or lacked phosphopeptide discrimination. Moreover, neither the third nor the fourth BRCT domain alone bound to phosphopeptides, which suggests that the tandem C-terminal BRCT domains function as a single module that is necessary and sufficient for phospho-specific binding.

BRCT domains are often found in tandem pairs [(BRCT)2] or multiple copies of tandem pairs, and the BRCA1 (BRCT)2 domains behave as a single stable fragment in limited proteolysis and x-ray crystallographic studies (13). Like those of PTIP, the (BRCT)2 domains of BRCA1, but not the individual BRCT domains alone, displayed phosphospecific binding (Fig. 2A). However, little if any phospho-binding was seen for (BRCT)2 modules from the human DNA damage-response proteins 53BP1 and MDC1, Saccharomyces cerevisiae Rad9p, or the PTIP N-terminal pair, which suggests that the phosphopeptide-binding function is present in only a subset of (BRCT)2 domains.

Fig. 2.

Comparison of the (BRCT)2 domains and determination of the optimal phosphopeptide-binding motifs. (A) PTIP, BRCA1, BRCA1 M1775R, MDC1, 53BP1, and Rad9p (BRCT)2 domains were assayed for binding as in Fig. 1A. The peptide libraries used were pSQ (Fig. 1A), pS = biotin-ZGZGGAXXXXpSXXXX-XAKKK; pSXXF = biotinZGZGGA XXXXpSXXFX-XAYKKK (pS, pSer; pT, pThr; Z, aminohexanoic acid; X, all amino acids except Cys). Domain boundaries: PTIP, as indicated in Fig. 1; BRCA1 BRCT1 and BRCT2, amino acids 1633 to 1863; BRCT1 alone, amino acids 1633 to 1740; BRCT2 alone, amino acids 1741 to 1863; MDC1, amino acids 1874 to 2089; 53BP1, amino acids 1622 to 1972; Rad9p, amino acids 985 to 1309. (B) Strong selection by the PTIP-(BRCT)2 and BRCA1-(BRCT)2 domains for Phe (F) at the (pSer/pThr)-Gln +3 position (7.0 or 7.5), respectively (table S1) (21). Bar graphs show the relative abundance of each amino acid at a given cycle of sequencing versus its abundance in the starting peptide library mixture (23).

We used oriented peptide library screening to determine the optimal phospho-binding motifs for the C-terminal (BRCT)2 domains of PTIP and BRCA1 (Fig. 2B) (table S1). PTIP-(BRCT)2 and BRCA-(BRCT)2 displayed strongest binding to similar but not identical motifs, with extremely strong selection in the (pSer/pThr) +3 position for aromatic and aliphatic residues, or aromatic residues, respectively. More moderate selection was also observed at other positions, particularly (pSer/pThr) +2 and +5. We defined an optimal (BRCT)2-binding peptide (BRCTide) as Tyr-Asp-Ile-(pSer/pThr)-Gln-Val-Phe-Pro-Phe and verified motif data with the use of a solid-phase array of immobilized phosphopeptides (fig. S1).

Isothermal titration calorimetry showed that the optimal pSer-containing peptide bound to the C-terminal (BRCT)2 of PTIP with a dissociation constant of 280 nM, and to the (BRCT)2 of BRCA1 with a dissociation constant of 400 nM (table S2). Substitution of pThr for pSer reduced the affinity of the peptide for PTIP-(BRCT)2, although binding was still observed. Substitution of pTyr, Ser, or Thr for pSer abrogated binding, which confirmed that (BRCT)2 modules are pSer- or pThr-specific (fig. S1).

After treatment of U2OS human osteosarcoma cells with 10 Gy of γ-irradiation, both PTIP-(BRCT)2 and BRCA1-(BRCT)2 bound distinct phosphoproteins, as detected by immunoblotting with a phospho-specific antibody to the (pSer/pThr)-Gln motif generated by ATM and ATR (Fig. 3, A, B, and D). Binding to (BRCT)2 was inhibited by incubation with specific phosphopeptides but not by nonphosphorylated peptides. Pretreatment of the cells with caffeine to inhibit the activity of ATM and ATR before irradiation also largely eliminated binding (14).

Fig. 3.

Association of PTIP and BRCA1 (BRCT)2 domains with DNA damage–induced phosphoproteins through their phosphopeptide-binding pockets. (A) Lysates from U2OS cells before or 2 hours after 10 Gyof γ-irradiation were incubated with GST-PTIP-(BRCT)2. Bound proteins were detected by immunoblotting with an antibody to the (pSer/pThr)-Gln motif generated by ATM and ATR. The interaction was disrupted by incubation with the pSQ peptide library, but not with the SQ peptide library or an unrelated pTP library (5). (B) Interaction of the PTIP-(BRCT)2 domain with phosphoproteins was disrupted by treatment of U2OS cells with caffeine (5 mM) before irradiation or by incubating the beads with BRCTtide (7pT), but not by its nonphosphorylated counterpart (7T). (C) (BRCT)2 domains of PTIP interact with 53BP1 after DNA damage. Endogenous 53BP1 from irradiated U2OS cells was precipitated with GST-PTIP-(BRCT)2 and detected by immunoblotting with an antibody to 53BP1. U2OS cells were transfected with HA-tagged 53BP1, and the interaction with GST-PTIP-(BRCT)2 was analyzed as in (A) and (B). Treatment of the cell lysates with lambda phosphatase also abolished the interaction. (D) Lysates from U2OS cells before or after irradiation were incubated with GST-BRCA1-(BRCT)2 and analyzed as in (B). In these experiments, the pSer version of BRCTtide (7pS) and its nonphosphorylated counterpart (7S) were used.

In response to γ-irradiation, 53BP1 undergoes multisite phosphorylation by ATM and forms nuclear foci (1517). We found that 53BP1 bound to the C-terminal PTIP-(BRCT)2 only after irradiation, and in a phosphorylation-dependent manner (Fig. 3C). In addition, in vivo association of full-length FLAG-tagged PTIP with hemagglutinin (HA)–tagged 53BP1 increased after irradiation (18).

The binding of (BRCT)2 domains to ATM- or ATR-phosphorylated substrates could localize PTIP to sites of DNA damage in vivo. In the absence of DNA damage, PTIP tagged with green fluorescent protein (GFP-PTIP) was diffusely nuclear with a small amount of cytosolic staining (Fig. 4). Two hours after irradiation, PTIP localized into discrete nuclear foci with (pSer/pThr)-Gln phosphoepitopes, 53BP1, and γ-H2AX (Fig. 4A). PTIP lacking the C-terminal (BRCT)2 did not form foci (Fig. 4B). The isolated C-terminal (BRCT)2 was diffusely nuclear in the absence of DNA damage, but relocalized into these nuclear foci after irradiation (Fig. 4C). Inhibition of ATM and ATR by caffeine before irradiation reduced the number and altered the appearance of full-length PTIP foci, and caused loss of colocalization with γ-H2AX (fig. S2). These findings strongly suggest that PTIP functions as a key component of the DNA damage response and may provide a molecular rationale for the early embryonic lethality of PTIP knockout mice with extensive unrepaired DNA ends (10).

Fig. 4.

Full-length PTIP forms DNA damage–induced foci and colocalizes with (pSer/pThr)-Gln proteins, 53BP1, and γ-H2AX. All cells shown were either treated with 10 Gy of γ-irradiation or mock-irradiated 24 hours after transfection, allowed to recover for 2 hours, stained, and analyzed by immunofluorescence microscopy. (A) U2OS cells were transfected with a full-length GFP-PTIP construct (PTIP-FL residues 1 to 757). (B) U2OS cells were transfected with a PTIP deletion construct in which the last two BRCT domains had been removed (PTIP-ΔBRCT, residues 1 to 549). (C) U2OS cells were transfected with a PTIP construct containing only the last two BRCT domains [(BRCT)2, residues 550 to 757].

(BRCT)2 selection for aromatic and aliphatic residues in the (pSer/pThr) +3 positions exceeds their selection for Gln in the +1 position. Thus, only a subset of ATM- and ATR-phosphorylated substrates are likely to bind with high affinity, and other kinases might also generate (BRCT)2-binding motifs. The important role for (BRCT)2 domains as pSer- or pThr-binding modules is emphasized by the finding that ∼80% of germline mutations in BRCA1 result in C-terminal truncations involving the BRCT region, predisposing women to breast and ovarian cancer (12). Interestingly, a BRCA1 cancer-associated mutation in the (BRCT)2 module that ablates critical BRCA1 protein interactions (19), Met1775 → Arg (M1775R), fails to bind phosphopeptides (Fig. 2A), even though the M1775R crystal structure is nearly identical to that of the wild-type (BRCT)2 (20). Agents that interfere with DNA damage signaling sensitize tumor cells to killing by radiation and chemotherapy. Thus, the phosphopeptide-binding pocket of (BRCT)2 modules may constitute a target for anticancer drug development.

Supporting Online Material

www.sciencemag.org/cgi/content/full/302/5645/636/DC1

Materials and Methods

Figs. S1 and S2

Tables S1 and S2

References

References and Notes

View Abstract

Navigate This Article