doi:10.1006/jmbi.2000.4291 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 304, 941±951 Structure of the FHA1 Domain of Yeast Rad53 and Identification of Binding Sites for both FHA1 and its Target Protein Rad9 Hua Liao, Chunhua Yuan, Mei-I Su, Suganya Yongkiettrakul, Dongyan Qin, Hongyuan Li, In-Ja L. Byeon, Dehua Pei* and Ming-Daw Tsai* Departments of Chemistry and Biochemistry, The Ohio State Biochemistry Program, and Campus Chemical Instrument Center, The Ohio State University, Columbus OH 43210, USA Forkhead-associated (FHA) domains have been shown to recognize both pThr and pTyr-peptides. The solution structures of the FHA2 domain of Rad53 from Saccharomyces cerevisiae, and its complex with a pTyr peptide, have been reported recently. We now report the solution structure of the other FHA domain of Rad53, FHA1 (residues 14-164), and identi®cation of binding sites of FHA1 and its target protein Rad9. The FHA1 structure consists of 11 b-strands, which form two large twisted anti-parallel bsheets folding into a b-sandwich. Three short a-helices were also identi®ed. The b-strands are linked by several loops and turns. These structural features of free FHA1 are similar to those of free FHA2, but there are signi®cant differences in the loops. Screening of a peptide library [XXX(pT)XXX] against FHA1 revealed an absolute requirement for Asp at the 3 position and a preference for Ala at the 2 position. These two criteria are met by a pThr motif 192TEAD195 in Rad9. Surface plasmon resonance analysis showed that a pThr peptide containing this motif, 188 SLEV(pT)EADATFVQ200 from Rad9, binds to FHA1 with a Kd value of 0.36 mM. Other peptides containing pTXXD sequences also bound to FHA1, but less tightly (Kd 4-70 mM). These results suggest that Thr192 of Rad9 is the likely phosphorylation site recognized by the FHA1 domain of Rad53. The tight-binding peptide was then used to identify residues of FHA1 involved in the interaction with the pThr peptide. The results are compared with the interactions between the FHA2 domain and a pTyr peptide derived from Rad9 reported previously. # 2000 Academic Press *Corresponding authors Keywords: FHA domain; Rad53; Rad9; phosphopeptide; NMR Introduction The forkhead-associated (FHA) domain was ®rst identi®ed as a 55-75 amino acid module found in many proteins from yeast to human, including nuclear protein kinases and transcriptional factors (Hofmann & Bucher, 1995). The conserved regions of FHA domains are not continuous, as three highly This is the third paper in the series about the structure-function relationship of FHA domains. For paper II, see Wang et al. (2000). {The ®rst four authors contributed equally to this work. Abbreviations used: AP, alkaline phosphatase; B, b-alanine; BCIP, 5-bromo-4-chloro-30 -indolylphosphate p-toluidine salt; FHA, forkhead-associated domain; GST, glutathione S-transferase; HBTU, 2-(1H-benzotriazole-1-yl)-1,1,3,3tetramethyluronium hexa¯uorophosphate; HBS, Hepes buffer saline; HBS-EP, Hepes buffer saline with EDTA and P20; HOBT, 1-hydroxybenzotriazole; HSQC, heteronuclear single-quantum coherence; MALDI-TOF, matrix-assisted laser desorption ionization-time of ¯ight; NHS, N-hydroxysulfosuccinimide; Nle, norleucine; NOE, nuclear Overhauser effect; NOESY, nuclear Overhauser enhancement correlated spectroscopy; ppm, parts per million; Rad9p, phosphorylated Rad9; TOCSY, total correlation spectroscopy; X, 19 amino acids including norleucine, except cystiene and methionine; Z, norleucine. E-mail addresses of corresponding authors: Pei.3@osu.edu; Tsai.7@osu.edu 0022-2836/00/050941±11 $35.00/0 # 2000 Academic Press 942 conserved blocks are separated with two segments with different lengths and sequences. Recently it has been shown that an additional ¯anking segment at each terminus is needed for the domain to be functionally active and structurally stable (Li et al., 1999; Liao et al., 1999; Hammet et al., 2000). Increasing evidence suggests that this domain participates in signal transduction pathways by promoting protein-protein interactions through recognition of phosphorylation sites. The degree of sequence homology is relatively low among different FHA domains (ca 20-30 %). Yeast Rad53, a checkpoint protein that prevents cell division when DNA is damaged or incompletely replicated, contains two highly divergent FHA domains: FHA1 in the N-terminal region and FHA2 in the C-terminal region. Sun et al. (1998) reported that the FHA2 domain of Rad53 from Saccharomyces cerevisiae interacts with phosphorylated Rad9 (abbreviated as Rad9p hereafter) in response to DNA damage at G2/M cell cycle phase, while Durocher et al. (1999) showed that both FHA domains of Rad53 interact directly with Rad9p. The latter paper also reported that a phosphothreonine peptide APPLSQE(pT)FSDLWKL derived from tumor-suppressor p53 can compete effectively for binding between FHA1 and Rad9p, but not binding between FHA2 and Rad9p. Using alanine scanning for the residues from 3 to 3, they demonstrated that the residue most critical for recognition is an Asp at 3 position. Together, these observations suggest that both FHA domains of Rad53 could interact with Rad9p but with distinct phosphopeptide-binding speci®city. We recently reported the ®rst structure of an FHA domain, the FHA2 domain of Rad53 (Liao et al., 1999). We also showed that a pTyr-containing peptide derived from Rad9, 826EDI(pY)YLD832, binds to FHA2 with modest af®nity (Kd ca 100 mM). The structure of the complex between FHA2 and this pTyr peptide has been subsequently solved (Wang et al., 2000). Furthermore, we have screened a pTyr peptide library, identi®ed consensus sequences for binding of pTyr peptides to FHA2, and demonstrated that two speci®c pTyr peptides bind to FHA2 tightly (Kd 1-5 mM). Taking the results from our laboratory and others together, we proposed that FHA domains could have pThr/Ser and pTyr dual speci®city, or that different FHA domains could have different phosphopeptide speci®city (Liao et al., 1999; Wang et al., 2000). It is important to note that none of the reports mentioned above have established the actual sites of Rad9 that are recognized by FHA1 or FHA2. The tight-binding pThr peptide identi®ed by Durocher et al. (1999) for binding to FHA1 was not related to Rad9. In our studies, FHA2 was only tested for binding of pTyr peptides. Thus, our results only suggested that, if the recognition site involves pTyr, the Tyr829 of Rad9 is the most likely candidate. As pointed out previously (Wang et al., 2000), our results could not rule out pThr or pSer as the natural recognition site of FHA2. Our NMR Structure and Speci®city of FHA1 preliminary results indicate that the FHA2 domain does bind to pThr peptides, but the binding af®nity and speci®city are still under investigation. The present paper focuses on the studies of structure and speci®city of the interaction between the FHA1 domain of Rad53 and phosphothreonine peptides. In what follows, we ®rst report the solution structure of FHA1 (residues 14-164), then present the use of mass spectrometry to identify selected pThr peptides from the screening of a pThr peptide library. On the basis of the consensus sequences obtained, we identi®ed a pThr peptide from Rad9 that binds tightly to FHA1. This peptide represents the most likely site of Rad9 that is recognized by the FHA1 domain of Rad53. This pThr peptide was then used to probe the residues of FHA1 that are involved in binding of the pThr peptide. Results and Discussion Construction and characterization of the FHA1 domain of Rad53 We started with the expression and puri®cation of fragment 2-175. The domain was expressed as a GST-fusion protein and then digested with thrombin. However, the product contained a mixture of two fragments of slightly different size due to a minor thrombin-cleavage site near residue 165. Subsequently, the fragment 2-164 was constructed. However, the ®rst 15 residues were shown to exist in random coil conformations by NMR. These residues were best deleted in order to avoid interfering with assignments. The fragment of residues 14-164 was ®nally constructed, which has been used for structural determination in this work. This version displayed an almost identical HSQC spectrum to those of longer versions, indicating retention of the overall structure. Figure 1 shows the amino acid sequences of FHA1 and FHA2 domains used in our structural studies. The sequence alignment shown in Figure 1 was assisted by structural alignment as described below. Determination of the structure of free FHA1 by NMR The chemical shift assignments were based on multidimensional NMR experiments such as 3D HNCACB, CBCA(CO)NH, HNCA, HCCH-TOCSY, and 15N-edited TOCSY and NOESY. All but six (K63, G79, N86, G94, D96, and G97) backbone amide groups have been assigned, and the chemical shift assignments of side-chains were more than 92 % complete. It is interesting to note that more than 38 % of Ha protons experience a >0.2 ppm down®eld shift from the respective random coil values, indicating rich b-strands in this protein. NOE assignments were performed on 3D 15 N-edited NOESY recorded in H2O, 3D 13C-edited NOESY recorded in 2H2O, and 2D homonuclear NOESY experiments recorded in H2O and 2H2O. 943 NMR Structure and Speci®city of FHA1 Figure 1. Sequences of the FHA1 domain used in this work and the FHA2 domain used in previous studies (Liao et al., 1999; Wang et al., 2000). The alignment is based on sequence homology as well as structural homology as explained in the text. Secondary structural elements are shown in blue for b-strands and pink for a-helices. Identical residues are indicated by stars. Hydrogen-deuterium exchange at 293 K reveals a set of slow exchanging amides, which together with NOE assignments helped to identify the secondary structure. An ensemble of 50 structures was generated with X-PLOR from 1962 distance constraints including 78 hydrogen bond restraints. A total of 192 backbone torsional angles f and were also included in the structural calculation, which were predicted from chemical shifts of 15N, 1 a 13 a 13 b H , C , C , and 13C0 with the program TALOS (Cornilescu et al., 1999). Some 20 resulting structures possessing low X-PLOR energy, as well as satisfying the experimental restraints (e.g. no NOE violations over 0.5 angstrom), were selected to represent the structure of FHA1 as shown in Figure 2(a). The average RMSD for this ensemble is Ê for backbone heavy atoms (N, Ca, 0.51(0.05) A Ê for all heavy atoms of and C) and 0.96(0.05) A residues 28-157 excluding the loop 52-62. The conformations of the N-terminal (residues 14-27) and the C-terminal tails (158-164) are not de®ned as these regions do not show long-range NOEs. The loop 52-62 is also not well de®ned compared with other regions, as a result of lacking long-range NOE. Overall, the structures display good stereochemical quality as evaluated by PROCHECK (Laskowski et al., 1993). The structural statistics are summarized in Table 1. Structural description of FHA1 As shown in Figure 2(b), the FHA1 structure consists of mainly anti-parallel b-strands, numbered from b1-b10 sequentially and b30 for the short strand that follows and parallels with b3. They form two large twisted anti-parallel b-sheets, which further fold into a b-sandwich. Three short a-helices were also identi®ed based on NOE assignments, one on each terminus (residues 16-24 and residues 149-157), and the third one in the loop that links b2 and b3 (residues 52-57). The b-strands are linked by several irregular loops and turns. The ®ve absolutely conserved residues in FHA domains are located in these loops and turns: G69 and R70 follows b3 immediately, S85 and H88 reside in the loop linking b30 and b4, and N107 is positioned in a turn immediately following b5. Similar to the case of FHA2 (Liao et al., 1999), the ``FHA1 domain'' de®ned by sequence analysis (residues 66-116) (Hofmann & Bucher, 1995) does not constitute a minimal structural domain. The ¯anking sequences are required to form a stable bsandwich structure. With the exception of several residues at N and C-termini, the current version of residues 14-164 appears to be suf®cient to maintain the structural integrity. Structural comparison between FHA1 and FHA2 Three structures in the Protein Data Bank were found to possess similar modular structure Table 1. Structural statistics of 20 FHA1 structures Restraint Long range NOE (ji jj 5 5) Medium range NOE (1 < ji jj < 5) Sequential NOE ((ji jj 1) Intraresidue NOE Hydrogen bond constraints Dihedral angles fa Dihedral angles a Ê) RMSD for distance restraints (A RMSD from idealized covalent geometry Ê) Bonds (A Angles (deg.) Impropers (deg.) PROCHECK (Ramachandran plot) Most favored regions (%) Additionally allowed region (%) Generously allowed region (%) Disallowed region (%) RMSD with respect to mean structure Backbone (residues 27-51, 63-157) Heavy atom (residues 27-51, 63-157) Backbone (residues 27-157) Heavy atom (residues 27-157) Number of Restraints 671 193 490 532 78 96 96 0.045 0.001 0.0035 0.0000 0.51 0.01 0.50 0.01 77.3 1.7 16.5 2.0 4.7 1.0 1.5 0.6 0.51 0.05 0.96 0.05 0.74 0.16 1.26 0.15 None of the 20 structures exhibited distance violations Ê. greater than 0.5 A a The torsion angle restraints were derived by using TALOS (Cornilescu et al., 1999). 944 NMR Structure and Speci®city of FHA1 Figure 2. (a) Stereoview showing superposition of peptide backbone traces (N, Ca C0 ) of 20 FHA1 structures (residues 27-158). Five residues are numbered for convenience of tracing the backbone. The N-terminal (14-26) and Cterminal (159-164) segments are disordered and thus not displayed. (b) Schematic ribbon diagram of a representative FHA1 structure. The 11 b-strands are numbered from 1 to 10 for the well de®ned anti-parallel strands and 30 for the marginally de®ned parallel b-strand (residues 76-77). This numbering system corresponds to the b-strands in FHA2 (Wang et al., 2000). The Figure was created with MOLMOL (Koradi et al., 1996). (b-sandwich) to FHA1 protein (Z-scores 57.2 for 1QU5, 1DEV, and 1YGS, whereas Z-score 43.0 for the rest) using the program DALI (Holm & Sanders, 1993), among which FHA2 (accession code 1QU5) has the closest fold (Z-score 8.5). Figure 3 shows the comparison of the ribbon diagrams between FHA1 and FHA2. Both domains consist of mainly b-strands (11 for FHA1 and 12 for FHA2). The 11 b-strands counted from the N terminus are structurally equivalent. The best superposition was achieved in the following segments: (1) residues 29-39 of FHA1 and residues 573-583 of FHA2, which include the b1 strand; (2) residues 40-50 of FHA1 and residues 586-596 of FHA2, which include the b2 strand; (3) residues 64-77 of FHA1 and residues 599-612 of FHA2, which include b3 and b30 strands; (4) residues 8296 of FHA1 and residues 616-630 of FHA2, which Figure 3. Stereoview of the superposition of the ribbon diagram structures of FHA1 (green, residues 26-155) and FHA2 (red, residues 570-702). The structures are superimposed according to the feedback from DALI (Holm & Sander, 1993). The segments 29-39, 40-50, 64-77, 82-96, 97-131, and 132-148 of FHA1 superpose well with the segments 573-583, 586-596, 599-612, 616-630, 645-679, 681-697, respectively, of FHA2. The residues that have large chemical shift perturbations upon peptide #2 binding (see Figure 7) are highlighted with green balls. The side-chains of R605, R617, and R620 in FHA2, which have been shown to interact directly with the phosphate group of a pTyr peptide, are shown in pink. The side-chains of the corresponding residues in FHA1, R70, R83, and K87, are shown in blue. The Figure was generated using MOLMOL (Koradi et al., 1996). 945 NMR Structure and Speci®city of FHA1 include the b4 strand; (5) residues 97-131 of FHA1 and residues 645-679 of FHA2, which include b5, b6, b7, b8, and b9 strands; (6) residues 132-148 of FHA1 and residues 681-697 of FHA2, which include the b10 strand. The RMSD of backbone atoms (N, Ca, C0 ) for these superimposed 103 resiÊ . Such structural alignments were dues is 2.45 A used along with sequence homology to achieve the sequence alignment between FHA1 and FHA2, as shown in Figure 1. As also shown in Figures 1 and 3, there are several noticeable differences in non-conserved regions. First, the lengths of the loops are different; the loop between b2 and b3 is short in FHA2 but longer in FHA1 (forming an a-helix), while the opposite is true for that between b4 and b5. Second, b11 and a long loop following b10 in FHA2 do not exist in FHA1. Consequently, the packing of the C-terminal helix against the top b-sheet does not occur in FHA1. In FHA1, a helix immediately follows b10. Comparison between the structures of FHA1 and FHA2 also allows us to de®ne further the ``minimal structural unit'' of an FHA domain. By eliminating the N-terminal and C-terminal segments that have random conformations or assume different secondary structures, it is possible that the minimal structural units for FHA1 and FHA2 domains are residues 29-148 and residues 573-697, respectively. However, whether these minimal units will be able to form a stable structure with full binding capability remains to be tested. Ligand specificity of FHA1 As the ®rst step toward identifying the natural ligand of FHA1, we performed systematic speci®city analyses for the FHA1 domain using a combinatorial pThr peptide library constructed on TentaGel S resin. The overall approach is similar to the screening of a pTyr peptide library for binding to FHA2 reported previously (Wang et al., 2000). The pThr library with random amino acid residues at 1 to 3 and 1 to 3 positions was designed: acetyl-AXXX(pT)XXXABBRM-resin. A peptide linker, ABBRM (B b-alanine), was synthesized to the C terminus of the random region (Yu & Chu, 1997). Each of the random positions has an equal representation by 18 natural amino acid residues (except for Cys and Met) plus norleucine (Nle), which is used as a substitute for Met. Therefore, the theoretical diversity of the library is 196 or 4.7 107. Each bead carries 100 pmol of a unique peptide sequence. The biotinylated FHA1 was used for pThr-library screening. The association of biotinylated FHA1 to resin-bound pThr peptide was detected by observing the blue color formation on the resin bead as a result of recruitment of SA-AP and subsequent enzymatic cleavage of BCIP. A negative control in which the reaction was performed without the biotinylated FHA1 yielded no color. Similar experiments were also performed for pSer and pTyr libraries (unpublished results). All three types of libraries gave blue beads, but the pThr library gave signi®cantly more beads with more intense color. Hence, in this work we focused on the identi®cation of consensus sequences for pThr peptides. Screening of 100 mg of resin (approximately 300,000 beads) resulted in 383 colored beads. Complete sequences were obtained for 117 beads by MALDI-TOF mass spectrometry. Figure 4 summarizes the frequencies of appearance for each residue at each of the randomized positions. The most striking feature of the results is that every selected sequence contains an Asp at the 3 position. Although Durocher et al. (1999) reported that replacement of Asp by Ala in the peptide APPLSQE(pT)FSDLWKL resulted in markedly decreased binding af®nity to FHA1, our results unequivocally show that the (pT)XXD is an absolute consensus sequence for high-af®nity binding to FHA1. Furthermore, our results further show that the 2 position prefers Ala (the highest abundance) and Ile (the second). It is important, however, to note that the (pT)XXD is not a universal recognition sequence for all FHA domains. Our unpublished results showed that the FHA2 domain of Rad53 also binds pT peptides, but with different consensus sequences (D is not preferred at the 3 position). Although the number of beads screened (3 105) represents only 0.6 % of the theoretical diversity of the library (4.7 107), we believe the results shown in Figure 4 are suf®cient to represent the ligand speci®city of FHA1 because the FHA domain-pThr peptide interaction clearly does not involve all six positions. The number of beads screened is suf®cient to cover any random tetrapeptide systems (194 1.3 105). This argument is supported by the results shown in Figure 4: some positions show only weak preferences, and position 3 shows little or no preference. For the positions that show strong preferences (positions 3 and 2), the results are very clear. Furthermore, our goal is not to identify a peptide that binds FHA1 most tightly in an absolute sense; instead, we intend to use the information to help us identify the site of Rad9 that is recognized by FHA1, described below. Identification of possible site(s) of Rad9 involved in binding with FHA1 Since the two most important features in the results of library screening is that the 3 position has an absolute requirement for Asp and the 2 position has a strong preference toward Ala, we searched the sequence of Rad9 for these features and found a match for residues 192TEAD195. We therefore obtained a pThr peptide containing this site, 188SLEV(pT)EADATFVQ200 (designated as peptide #2) for analyses of its binding to FHA1. In addition to the best match for positions 3 and 946 NMR Structure and Speci®city of FHA1 Figure 4. Results of pThr peptide sequence speci®city for binding to the FHA1 domain. Displayed are the amino acid residues selected at 3 to 3 positions relative to the pThr residue. The x-axis indicates the identity of the 19 amino acid residues in single-letter codes, whereas abundance on the y-axis represents the number of occurrence of an amino acid residue at a certain position. M, norleucine. 2, this peptide also has a reasonably good match at other positions: position 2 (E) is also the best ®t, while position 1 (V) and position 1 (E) are among the most favorable residues. Thus, on the basis of sequence analyses, we predicted that this peptide of Rad9 is likely to contain the main residues recognized by FHA1 and thus binds tightly to FHA1. For the purpose of comparison, the binding assays also included the four other peptides containing TXXD sites from Rad9: 148KKMTFQ(pT)PTDPLE160 (peptide #1), 537 SNAA(pT)RDDIIIAG549 (peptide #3), and 803 LEVD(pT)NQDGCWIY815 (peptide #4), and 1293 HDDI(pT)DNDI1301 (peptide #5). In addition, peptide #6, SGSYSQE(pT)FSDLG, was designed on the basis of the results by Durocher et al. (1999). The residues from 3 to 4 in this peptide are identical to a peptide reported by Durocher et al., APPLSQE(pT)FSDLWKL, which was shown to bind tightly to GST-FHA1 with Kd 1.68 mM as determined by isothermal titration calorimetry. The dissociation constants of the FHA1-pThr peptide complexes were determined by surface plasmon resonance (SPR), which is a very useful technique in characterizing biomolecular interactions (Karlsson & Stahlberg, 1995; Lookene et al., 2000; Schuck, 1997). Figure 5(a) shows the sensorgrams of FHA1 binding to peptide #2. The Kd value (0.36 ( 0.03) mM) was obtained by ®tting the maximal response at each FHA1 concentration to a hyperbola curve (Figure 5(b)). The Kd values of all pTXXD peptides are summarized in Table 2, which indicate that all six peptides bind to FHA1 with reasonably good af®nity, since they all satisfy the absolute requirement of pTXXD. However, the binding af®nity of peptide #2 is substantially higher than that of the rest of the peptides. These results support that the most likely residues of Rad9 involved in the binding of FHA1 are Thr192 (in the phosphorylated state) and the surrounding residues. However, it cannot be ruled out that FHA1 binds to one or more of the other pTXXD sites of Rad9. We further hypothesize that, while the absolutely required Asp residue at 3 probably contributes signi®cantly to the binding energy, the residues at other positions could contribute to the speci®city of binding between different FHA domains and their speci®c targets. This hypothesis will be tested by further structural and functional analyses. Although the Kd value of peptide #6 (13.6 mM) is higher than the value of 1.68 mM (measured by isothermal titration calorimetry) reported by Table 2. Dissociation constants of FHA1-pThr peptide complexes Peptide # #1 #2 #3 #4 #5 #6 Sequence 148 Kd (mM) 160 KKMTFQ(pT)PTDPLE SLEV(pT)EADATFVQ200 537 SNAA(pT)RDDIIIAG549 803 LEVD(pT)NQDGCWIY815 1293 HDDI(pT)DNDI1301 SGSYSQE(pT)FSDLG 188 10.2 0.3 0.36 0.03 71.1 2.2 3.5 0.3 5.03 0.19 13.6 0.8 947 NMR Structure and Speci®city of FHA1 Identification of possible residues of FHA1 involved in binding with Rad9 Figure 5. (a) Surface plasmon resonance sensorgrams for the binding of FHA1 protein to immobilized pThr peptide #2, 148KKMTFQ(pT)PTDPLE160, derived from Rad9. The control signals, which were generated by ¯owing FHA1 protein through a blank channel without peptide bound on the surface, have already been subtracted. (b) Plots of the maximal responses in the sensorgrams versus the concentration of FHA1. The Kd value was extracted by ®tting the maximal response at each FHA1 concentration to a hyperbola curve y a*x/ (Kd x), where y is the response in RU (resonance unit, in arbitrary value), x is the protein concentration, and a is the maximum response. Durocher et al. (1999), these two peptides are not identical, nor are the FHA1 domains used in our study (ours is shorter and theirs is longer and with a GST tag). Another point which should be noted in comparing our results with their results is that they have obtained Kd values in the order of 50100 nM using BIAcore, but as explained in their paper, these lower Kd values were caused by GSTmediated protein dimerization effects in the SPR analysis. Since our FHA1 domain was not fused to GST, such a problem did not exist in our analysis. In any case, the Kd values obtained from SPR analyses are best used in comparing relative binding af®nities between different ligands, which is the main purpose of our studies. Above we identi®ed possible residues of Rad9 involved in binding with FHA1. The goal of this section is to obtain complementary information: possible residues of FHA1 involved in binding with Rad9. Our approach is to use 15N-HSQC experiments to identify the residues of FHA1 that are perturbed upon binding of peptide #2. For the purpose of comparison, pThr peptide #1 was also included in such studies. The results of titration experiments, as shown in Figure 6, indicate that FHA1-peptide #1 are in ``fast exchange'' on the NMR time scale, whereas FHA1-peptide #2 are in ``slow exchange'', which is clear evidence that peptide #2 binds to FHA1 more tightly than peptide #1. NMR is not a sensitive method for quantitative Kd measurements. However, NMR results support the results of binding analyses by SPR. The following FHA1 residues are perturbed by >0.05 ppm for protons or >0.4 ppm for nitrogen atoms upon binding peptide #2: T67, F68, G69, R70, N71, I81, S82, S85, K87, T106, N107, K118, G133, V134, G135, V136, and E137. A subset of these residues (those with >0.1 ppm shifts in proton or >0.6 ppm shifts in 15N) are shown in Figure 7 and highlighted in Figure 3. The ®rst three residues are located in the b3-strand. The rest are in the loops or turns: R70 and N71 in the b3-b30 loop, I81, S82, S85, and K94 in the b30 -b4 loop, T106 and N107 in the b5-b6 loop, K118 in the b7b8 loop, and G133-E137 in the b9-b10 loop. These residues are the likely (but not the only) candidates for interacting with the pThr peptide. As shown in Figure 3, these residues reside in the front-bottom pocket formed by the loops. It is worthwhile to point out that the ¯anking sequences could also contribute to binding of the phosphopeptide, in addition to their contribution to the structural integrity. Similar results were obtained for the pThr peptide #1 (Figure 7) and peptide #5 and #6 (not shown), though the binding af®nities and overall chemical shift perturbations are generally smaller relative to those of peptide #2. Detailed structural analyses of these complexes are in progress, which is required to understand fully the structural basis of ligand speci®city. Comparison between FHA1-pThr peptide and FHA2-pTyr peptide Here we compare the result of FHA1-pThr peptide complex presented in this paper with that of FHA2-pTyr peptide complex reported previously (Liao et al., 1999; Wang et al., 2000). Our results suggest that the same set of structurally equivalent residues are involved in the phosphopeptide binding in both complexes. For example, N71, S85 and G133 in FHA1 are aligned structurally with S606, S619 and W682, respectively, of FHA2 (Figure 1), and all of these residues display substantial chemi- 948 NMR Structure and Speci®city of FHA1 Figure 6. Partial 2D 1H-15N HSQC spectra showing fast exchange and slow exchange when FHA1 is titrated with peptide #1 and #2, respectively. (a) Overlay of partial HSQC spectra of free FHA1 (black) and FHA1 in the presence of 0.5, 1.0, 2.0 equivalents (red, green, and blue, respectively) of peptide #1, with arrows showing the direction of shifts. (b) Single HSQC spectrum of FHA1 in the presence of ca 0.3 equivalents of peptide #2, which reveals two sets of FHA1 amide peaks in slow exchange. Binding experiments were performed by adding aliquots of the peptide to a sample of uniformly 15N-labeled FHA1 in 10 mM phosphate buffer, pH 6.5 at 293 K. Experiments were performed on a Bruker DRX-800 NMR spectrometer at 293 K. cal shift perturbations in both complexes. However, the relative magnitudes of perturbations differ, suggesting differences in detailed interactions. In the FHA2-pTyr peptide complex, Arg605, Arg617, and Arg620 are the three key residues interacting with the phosphate group. These three residues correspond to Arg70, Arg83, and Asn86 of FHA1 in the sequence alignment shown in Figure 1. The two arginine residues have been perturbed seriously in side-chain resonances, although side-chain assignments are tentative at this stage. Asn 86, or its neighboring residue Lys87, could be the third residue involved in phosphate binding. As shown in Figure 7, Lys87 is one of the residues with largest chemical shift perturbations. The sidechains of Arg70, Arg83, and Lys87 of FHA1 are highlighted in Figure 3. Although they do not superpose well with the side-chains of Arg605, Arg617, and Arg620 of FHA2, they are well positioned to bind the phosphate group. This comparison suggests that the pThr-peptide binding site of FHA1 is generally similar to the pTyr-peptide binding site of FHA2, but there are clear differences in detailed interactions that can only be elucidated when high-resolution structures of FHA1pThr peptide complexes become available. However, the results and discussions presented in this paper, along with our unpublished ®nding that the FHA2 domain of Rad53 also binds some pThr peptides tightly (but with different consensus sequences), lends further support to our previous proposal that FHA domains can bind to both pThr and pTyr peptides chemically (Liao et al., 1999; Wang et al., 2000). Biologically, whether the FHA2 domain of Rad53 binds to a pThr site or a pTyr site of Rad9 remains to be established, but the results presented here are suf®cient to suggest that the FHA1 domain of Rad53 very probably binds to Thr192 and surrounding residues of Rad9. Materials and Methods Materials Figure 7. Summary of FHA1 residues whose 15N or H chemical shifts are perturbed by binding of peptide #1 (two equivalents) and peptide #2. The y-axis indicates the magnitudes of shifts. Only the residues perturbed by >0.10 ppm for 1H chemical shifts or >0.6 ppm for 15N chemical shifts upon binding of peptide #2 are shown. Three other residues (R70, T106 and E137), for which only lower limit can be estimated due to incomplete assignments, could also belong to this category. 1 TentaGel S NH2 resin, N-9-¯uorenylmethoxycarbonyl (Fmoc)-amino acids, 1-hydroxybenzotriazole (HOBt) and 2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexa¯uorophosphate (HBTU) were purchased from Advanced ChemTech (Louisville, KY). N-Hydroxysuccinimidobiotin, acetylglycine, N-acetyl-D,L-alanine, 5bromo-4-chloro-3-indolyl phosphate, and streptavidinalkaline phosphatase were obtained from Sigma Chemical Co. All other chemicals were purchased from Aldrich Chemical unless stated otherwise. Protein expression and purification Three versions of FHA1 domain were constructed as fragments 2-175, 2-164, and 14-164 of Rad53. They were cloned into a pGEX-4T vector (Pharmacia Biotech) for expression and puri®cation as GST fusion proteins. 949 NMR Structure and Speci®city of FHA1 Escherichia coli BL21(DE3) cells transformed with the plasmids were grown to mid-log phase in rich media. Expression was induced at 30 C by adding 1.0 mM IPTG. Cells were harvested eight hours later and lysed using lysozyme and sonication. GST fusion proteins were puri®ed using glutathione agarose (Sigma). The GST tag was removed by thrombin (Sigma) digestion and gel ®ltration chromatography. The 15N and 15N/13C isotopically labeled proteins were expressed in M9 minimal media with 15NH4Cl as the only nitrogen source and 13 D-glucose- C6 as the only carbon source. Protein puri®cation was checked by SDS-PAGE, and protein concentrations were determined spectrophotometrically at 280 nm using extinction coef®cient of 12900 M 1 cm 1. NMR spectroscopy and assignments NMR samples typically contained 0.5 mM protein, 10 mM sodium phosphate buffer, 1 mM DTT, and 1 mM EDTA in 90 % H2O/10 % 2H2O or 100 % 2H2O at pH 6.5. Data were acquired at 293 K on Bruker DRX-800 and DMX-600 spectrometers, both equipped with tripleresonance probes and three-axis gradients. The triple resonance experiments HNCACB (Muhandiram & Kay, 1994), CBCA(CO)NH (Grzesiek & Bax, 1992b), HNCA (Grzesiek & Bax, 1992a; Ikura et al., 1990), and HNCO (Grzesiek & Bax, 1992a; Ikura et al., 1990) as well as HCCH-TOCSY were performed on the DMX-600, while the NOESY-type experiments (e.g. 13C-edited NOESY and 15N-edited NOESY) and 3D 15N-edited TOCSYHSQC were carried out on the DRX-800. Mixing times were 100 ms and 30-40 ms for NOESY and TOCSY experiments. Data were processed using XWINNMR (Bruker) or Felix (Molecular Simulations Inc.) installed on Silicon Graphics workstations. Generally, one time zero-®lling was employed in all indirectly detected dimensions. Appropriate window function was applied on each dimension followed by Fourier transformation and baseline correction. Spectra were referenced indirectly to TSP at 0.00 ppm. Structure calculation The interproton NOE cross-peaks of 2D and 3D NOESY spectra were classi®ed as 1.8-2.8, 1.8-3.5, 1.8-5.0, Ê corresponding to strong, medium, weak, and 1.8-6.0 A and very weak NOEs. Upper limits for distances involving methyl protons and non-stereospeci®cally assigned protons were corrected appropriately for center averaging (Wuthrich et al., 1983). Hydrogen bond restraints Ê and 1.5-2.8 A Ê for N-O and H-O were added as 2.4-3.5 A internuclear distances, respectively, in regions where regular secondary structure elements were identi®ed by NOE as well as hydrogen-deuterium exchange results. The structures were calculated by using simulated annealing protocol implemented in the X-PLOR program, version 3.843 (Brunger, 1993; Nilges et al., 1988), installed on a Silicon Graphics computer. The quality of the NMR structure was externally viewed by MOLMOL (Koradi et al., 1996) and Insight II (Molecular Simulations Inc.) and evaluated by the PROCHECK program (Laskowski et al., 1993). The result was gradually improved by using an iterative strategy of structurebased assignment. The ®nal 20 structures with lowest energy among the converged structures were selected. Synthesis of peptide libraries and peptides The pTyr library used was the same as that described (Wang et al., 2000). The library of acetyl-AXXX(pT)XXXABBRM-resin and the corresponding pS library were constructed on an in-house built peptide synthesis apparatus. The synthesis was performed at room temperature on a 3.0 g scale of TentaGel S-NH2 resin (0.3 mmol/g loading, 2.86 106 beads/g, 80-100 mm) by using Fmoc/HBTu/HOBt chemistry (Bodanszky, 1993). The amino acid linker, ABBRM, was ®rst synthesized on the resin, followed by acetyl-AXXX(pT)XXX in which a split synthesis approach was used for randomization (Lam et al., 1991; Yu & Chu, 1997), resulting in a onebead, one-peptide library. Amino acid residues with the same molecular mass were differentiated by capping reagents: 5 % acetylalanine was added to the coupling reaction of Leu, while 5 % acetylglycine was added in the coupling reactions of Lys and norleucine. The complete coupling reaction of each round was monitored by ninhydrin test. Deprotection was performed in the ®nal stage by reacting for two hours with large excess of the solution made up of crystalline phenol, tri¯uoroacetic acid (TFA), anisole, thioanisole, and 1,2-ethanedithiol. The beads were then thoroughly washed with alternating methanol and dichloromethane, dried under vacuum, and kept at 20 C. Screening of resin-bound peptide libraries FHA1 was ®rst treated with sulfo-NHS-biotin (Pierce) to give biotinylated FHA1. Sulfo-NHS-biotin (in 1.2-fold excess) was mixed with the puri®ed FHA1 in 10 mM phosphate buffer, pH 7.5 for four hours at room temperature. The unbound biotin was removed from the reaction by dialysis. The biotinylated FHA1 was used for screening peptide libraries. TentaGel beads (100 mg) containing peptide libraries (approximately 300,000 beads) transferred to a polypropylene column connected to vacuum manifold (Bio-Rad) were sequentially washed with 100 ml HBS buffer (25 mM Hepes and 150 mM NaCl, pH 7.5) and 100 mL HBST buffer (HBS containing 0.1 % (v/v) Tween-20, pH 7.5). Subsequently, the beads were incubated with 20 ml HBSTB solution (HBST containing 1 mg/ml bovine serum albumin, pH 7.5) for one hour with mixing. After the incubation, the beads were washed twice with 50 ml HBSTB solution. The beads were then incubated with 0.1 mM biotinylated FHA1 in 20 ml HBSTB at room temperature for six to eight hours with mixing. The unbound proteins were then removed by washing with 100 ml HBSTB solution. The binding of biotinylated FHA1 to a peptide was observed by a staining procedure in which 0.5 mg/ml streptavidin-AP and 20 mM phosphate in 20 mL HBSTB were subsequently incubated with the beads for 15 minutes at room temperature. After incubation, the unbound streptavidin-AP was removed and the beads were washed with 100 ml HBST buffer, 150 ml HBS buffer and 100 ml TBS buffer (25 mM Tris-HCl and 150 mM NaCl (pH 8.0)). 0.5 mg/ ml BCIP in 20 ml TBS buffer (pH 8.0) was added into the reaction. The blue color typically developed on positive beads within ten minutes to one hour of incubation. The color intensity depends on the binding af®nity, with more intense color developed on beads that carry tighter binding sequences. The positive beads were selected from the library and subjected to peptide ladder sequencing as described (Hu et al., 1999; Wang et al., 2000; Youngquist et al., 1995), except that the peptide ladder was generated after screening using a novel partial 950 Edman degradation strategy (P. Wang and D. Pei, in preparation). The mass spectra were acquired in positive ion mode by MALDI-TOF mass spectrometer (Re¯ex III, Bruker Daltonics). Binding analysis by surface plasmon resonance The peptides were purchased from Genemed Sythesis, Inc., CA. Each peptide showed a single peak on analytical reversed-phase HPLC and gave a mass identical to that calculated. The dissociation constants between FHA1 (residues 2-175 with a V5 tag, which is longer than the fragment used in NMR studies) and the peptides were determined on a BIAcore 2000 instrument (Karlsson & Stahlberg, 1995) at 25 C. The target peptides were ®rst biotinylated by treating with sulfo-NHS-biotin (Pierce) on ice for two hours, followed by diluting in HBS-EP buffer (10 mM Hepes, 150 mM NaCl, 3 mM EDTA, 0.005 % Tween-20 (pH 7.4)) to a concentration of ca 2 mM. The SA sensor chip, with streptavidin from Streptomyces adidinii bound on the gold-dextran surface, was conditioned according to the manufacturer with one-minute injection of 1 M NaCl in 50 mM NaOH, three times in succession. The chip surface was regenerated with MgCl2 and NaCl after each individual measurement. A total of 40 ml peptide solution was injected each time at a rate of 10 ml/min. After extensive washing with HBS-EP buffer and a one-minute injection of 1 M NaCl to remove non-speci®cally bound peptides, a baseline increase of 100-200 resonance units (RU) is typically obtained. FHA1 solutions of increasing concentration were then ¯owed across the SA chip at a rate of 10 ml/min for 240 s. As a result of binding between pThr peptide and FHA1, changes build up with time forming the ``sensorgram''. The maximal response units at equilibrium binding condition are proportional to the mass of protein bound to the surface. At least ®ve different protein concentrations were performed and excess protein solutions were injected each time to reach the ``equilibrium region'' in the sensorgram, de®ned as the maximal response. The latter (after correcting for the control signals from the blank channel without immobilized peptide) was then plotted against the protein concentration to determine the equilibrium dissociation constant (Kd). Accession number Coordinates for the structure of free FHA1 have been deposited in the RCSB PDB (accession number 1G3G). Acknowledgments The authors thank Ohio Supercomputer Center for the T94 Cray supercomputer resource grant, Ohio Board of Regents for funding of the 800 MHz NMR and the mass spectrometers, and NIH for ®nancial support (CA87031). References Bodanszky, M. (1993). Principles of Peptide Synthesis, 2nd rev. edit., Springer-Verlag, Berlin, New York. Brunger, A. T. (1993). X-PLOR Version 3.1 manual, Yale University, New Haven, Connecticut. NMR Structure and Speci®city of FHA1 Cornilescu, G., Delaglio, F. & Bax, A. (1999). Protein backbone angle restraints from searching a database for chemical shift and sequence homology. J. Biomol. NMR, 13, 289-302. Durocher, D., Henckel, J., Fersht, A. R. & Jackson, S. P. (1999). The FHA domain is a modular phosphopeptide recognition motif. Mol. Cell, 4, 387-394. Grzesiek, S. & Bax, A. (1992a). Improved 3D tripleresonance NMR techniques applied to a 31-kDa protein. J. Magn. Reson. 96, 432-440. Grzesiek, S. & Bax, A. (1992b). Correlation backbone amide and side-chain resonances in larger proteins by multiple relayed triple resonance NMR. J. Am. Chem. Soc. 114, 6291-6293. Hammet, A., Pike, B. L., Mitchelhill, K. I., Teh, T., Kobe, B., House, C. M., Kemp, B. E. & Heierhorst, J. (2000). FHA domain boundaries of the dun1p and rad53p cell cycle checkpoint kinases. FEBS Letters, 471, 141-146. Hofmann, K. & Bucher, P. (1995). The FHA domain: a putative nuclear signalling domain found in protein kinases and transcription factors S. Trends Biochem. Sci. 20, 347-349. Holm, L. & Sanders, C. (1993). Protein structure comparison by alignment of distance matrices. J. Mol. Biol., 233, 123-138. Hu, Y. J., Wei, Y., Zhou, Y., Rajagopalan, P. T. R. & Pei, D. (1999). Determination of substrate speci®city for peptide deformylase through the screening of a combinatorial peptide library. Biochemistry, 38, 643650. Ikura, M., Kay, L. E. & Bax, A. (1990). A novel approach for sequential assignment of 1H, 13C, and 15N spectra of proteins: heteronuclear triple-resonance three-dimensional NMR spectroscopy. Application to calmodulin. Biochemistry, 29, 4659-4667. Karlsson, R. & Stahlberg, R. (1995). Surface plasmon resonance detection and multispot sensing for direct monitoring of interactions involving low-molecularweight analytes and for determination of low af®nities. Anal. Biochem. 228, 274-280. Koradi, R., Billeter, M. & Wuthrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graphics, 14, 2932, 51-55. Lam, K. S., Salmon, S. E., Hersh, E. M., Hruby, V. J., Kazmierski, W. M. & Knapp, R. J. (1991). A new type of synthetic peptide library for identifying ligand-binding activity [published errata appear in Nature 1992 Jul 30; 358(6385):434 and 1992 Dec 2431;360(6406):768]. Nature, 354, 82-84. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK - a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283-291. Li, J., Smith, G. P. & Walker, J. C. (1999). Kinase interaction domain of kinase-associated protein phosphatase, a phosphoprotein-binding domain. Proc. Natl Acad. Sci. USA, 96, 7821-7826. Liao, H., Byeon, I. J. L. & Tsai, M. D. (1999). Structure and function of a new phosphopeptide-binding domain containing the FHA2 of Rad53. J. Mol. Biol. 294, 1041-1049. Lookene, A., Stenlund, P. & Tibell, L. A. (2000). Characterization of heparin binding of human extracellular superoxide dismutase. Biochemistry, 39, 230-236. Muhandiram, D. R. & Kay, L. E. (1994). Gradientenhanced triple-resonance 3-dimensional NMR NMR Structure and Speci®city of FHA1 experiments with improved sensitivity. J. Magn. Reson. 103, 203-216. Nilges, M., Clore, G. M. & Gronenborn, A. M. (1988). Determination of three-dimensional structures of proteins from interproton distance data by dynamical simulated annealing from a random array of atoms. Circumventing problems associated with folding. FEBS Letters, 239, 129-136. Schuck, P. (1997). Use of surface plasmon resonance to probe the equilibrium and dyanmic aspects of interactions between biological macromolecules. Annu. Rev. Biophys. Biolol. Struct. 26, 541-566. Sun, Z., Hsiao, J., Fay, D. S. & Stern, D. F. (1998). Rad53 FHA domain associated with phosphorylated Rad9 in the DNA damage checkpoint [see comments]. Science, 281, 272-274. Wang, P., Byeon, I.-J. L., Liao, H., Beebe, K., Pei, D. & Tsai, M.-D. (2000). Structure and speci®city of the 951 interaction between the FHA2 domain of Rad53 and phosphotyrosine-peptides. J. Mol. Biol. 302, 927940. Wuthrich, K., Billeter, M. & Braun, W. (1983). Pseudostructures for the 20 common amino acids for use in studies of protein conformations by measurements of intramolecular proton-proton distance constraints with nuclear magnetic resonances. J. Mol. Biol. 169, 949-961. Youngquist, R. S., Fuentes, G. R., Lacey, M. P. & Keough, T. (1995). Generation and screening of combinatorial peptide libraries designed for rapid sequencing by mass spectrometry. J. Am. Chem. Soc. 117, 3900-3906. Yu, Z. & Chu, Y. H. (1997). Combinatorial epitope search: pitfalls of library design. Bioorg. Med. Chem. Letters, 7, 95-98. Edited by M. F. Summers (Received 20 September 2000; received in revised form 26 October 2000; accepted 26 October 2000)