ReviewBiochemistry

Structure and Function of the Phosphothreonine-Specific FHA Domain

See allHide authors and affiliations

Science Signaling  23 Dec 2008:
Vol. 1, Issue 51, pp. re12
DOI: 10.1126/scisignal.151re12

This article has a correction. Please see:

Abstract

The forkhead-associated (FHA) domain is the only known phosphoprotein-binding domain that specifically recognizes phosphothreonine (pThr) residues, distinguishing them from phosphoserine (pSer) residues. In contrast to its very strict specificity toward pThr, the FHA domain recognizes very diverse patterns in the residues surrounding the pThr residue. For example, the FHA domain of Ki67, a protein associated with cellular proliferation, binds to an extended target surface involving residues remote from the pThr, whereas the FHA domain of Dun1, a DNA damage–response kinase, specifically recognizes a doubly phosphorylated Thr-Gln (TQ) cluster by virtue of its possessing two pThr-binding sites. The FHA domain exists in various proteins with diverse functions and is particularly prevalent among proteins involved in the DNA damage response. Despite a very short history, a number of unique structural and functional properties of the FHA domain have been uncovered. This review highlights the diversity of biological functions of the FHA domain–containing proteins and the structural bases for the novel binding specificities and multiple binding modes of FHA domains.

Introduction

The forkhead-associated (FHA) domain, a protein-phosphoprotein interaction motif with high specificity for phosphothreonine (pThr) residues, has been identified in more than 2000 proteins (from the Pfam database) in prokaryotes and eukaryotes since its discovery in forkhead family transcription factors in 1995 (1). It is present in many regulatory proteins, kinases, phosphatases, and transcription factors (1) (Fig. 1A). It is now clear that FHA domains play important roles in human diseases, particularly in relation to DNA damage responses and cancers, and in biological processes such as cell growth, signal transduction, and cell cycle regulation (Fig. 2). Specific roles for FHA domains in checkpoint pathways have been explored with regard to various kinases in different species (Fig. 1B). Some examples of FHA domain–containing proteins include the human tumor suppressor checkpoint kinase 2 (Chk2), checkpoint with forkhead and ring finger domains (CHFR) (2), TIFA [tumor necrosis factor receptor (TNFR)–associated factor (TRAF)–interacting protein with a forkhead-associated domain] (3), and Nijmegen breakage syndrome 1 (NBS1, also known as nibrin) (4).

Fig. 1

Proteins with FHA domains. (A) Proteins with diverse functions in different organisms. (B) Proteins in the Ser-Thr kinase family that have functions specific to checkpoint pathways in different organisms. Other domains indicated are phosphatase, RING (ring finger), SCD (SQ-TQ cluster), PST (Pro-Ser-Thr), BRCT (breast cancer susceptibility gene–1 C-terminal), TB (TRAF6 binding), BTA (bacterial transcriptional activation), DBD (DNA-binding), and kinase domains. Numbers in parentheses indicate the total number of amino acid residues in each protein. [Information about sequences and domains to help prepare this figure was obtained from http://www.sanger.ac.uk/software/pfam/.]

Fig. 2

Diagram of proposed functions of FHA domain–containing proteins. See main text for details.

The FHA domain is unique among signal transduction domains in two aspects. First, it is the only presently known protein domain that specifically recognizes pThr; other domains such as WW and 14-3-3 recognize both pThr and pSer residues. Second, the FHA domain shows very diverse ligand specificity; different FHA domains recognize the pTXXD motif, the pTXXI/L motif, and TQ clusters (singly and multiply phosphorylated) (5). The Ki67-FHA even recognizes an extended binding surface, but not short phosphopeptides. Here, we discuss FHA domains in terms of their structure, binding specificity, and biological function, and we highlight multiple binding modes, possible alternative binding sites, and their importance in providing diversity to their functions in regulating multiple signaling pathways. Although the FHA domain has been known for only a relatively short period of time, a great deal of information has already been acquired, particularly since it was last reviewed in 2002 by Durocher and Jackson (6) and in 2003 by Heierhorst and colleagues (7).

Structures of Free FHA Domains

To date, the structures of 16 FHA domains have been determined either in their free forms, in complexes with ligand, or both, by nuclear magnetic resonance (NMR) and x-ray crystallography studies (http://www.rcsb.org) (Table 1). The first reported structure is the C-terminal FHA domain of the protein Rad53 identified from radiosensitive mutants of budding yeast, Rad53-FHA2, which was solved by NMR (8) (Fig. 3). Consistent with proteolytic digestion studies (9), all structures clearly show that the FHA domain is considerably larger than the 55 to 75 amino acid residues—the core FHA domain homology region—predicted by sequence analyses (1). The minimal structure of an FHA domain encompasses the essential 11 β strands and spans approximately from 95 residues (in the case of Ki67-FHA) to 121 residues (in the case of Rad53-FHA2). Although some N-terminal or C-terminal flanking sequences are often needed to obtain soluble and stable proteins, an FHA domain construct is still generally under 150 residues, which makes it well suited for structural determination by solution NMR studies.

Table 1 Structures of free FHA and FHA domain–phosphopeptide complexes. Conserved residues are underlined.
View this table:
Fig. 3

Basic structural features of the FHA domain. Stereoview of the FHA2 domain of the S. cerevisiae kinase Rad53. The large β sheets are colored in green and blue, respectively, and the conserved residues highlighted in the sequence alignment are labeled. Note that in some studies the very short and parallel β4 strand was labeled as β3′ and the subsequent strands as β4-β10 instead of β5-β11.

The FHA domain is typically characterized by the following features (Fig. 3): (i) It is rich in β strands; there are 11 more or less well-defined β strands numbered from β1 to β11. (ii) It consists of two large β sheets. One sheet contains six antiparallel strands (β2, β1, β11, β10, β7, and β8); the other sheet contains five mixed β strands (β4, β3, β5, β6, and β9), of which β4 is the shortest and is parallel to β3. (iii) It contains a twisted β sandwich; this is the core architecture of the tertiary fold formed by burying hydrophobic side chains between the two large β sheets. (iv) It has structural modularity; the first and last β strands intimately interact with each other as part of one piece of the β sandwich, and the N terminus and C terminus meet at the end of the domain opposite to the pThr-binding site (Fig. 3). In conclusion, despite sharing low sequence homology, all known FHA domains adopt a strikingly similar fold, a structural feature that could be attributed to the preservation of hydrophobicity in the β strands (10). A structure-based sequence alignment of FHA domains is shown in Fig. 4A.

Fig. 4

Sequence alignments. (A) FHA domains. The residues colored in red highlight the 11 conserved β strands among FHA domains, while yellow highlights indicate conserved residues. Dun1, FHA1, and FHA2 are from S. cerevisiae, and Cds1 is from S. pombe. Chk2, RNF8, and Ki67 are human proteins. All FHA domain alignments are based on their structures except for that of Cds1-FHA, which is aligned on the basis of its sequence. (B) SCD domains. Proteins with SCD domains that have been discussed in this review are presented. There is no evident homology among these proteins, but SQ and TQ residues are highlighted in red to illustrate their proximity to each other in a short stretch of protein. The numbers of amino acid residues omitted are shown between the SQ-TQ repeats. Sc, S. cerevisiae; Sp, S. pombe.

Major structural differences between different FHA domains occur in the loops and turns that connect the β strands, which can either vary substantially in length, contain a small α-helical insertion, or both. This is exemplified by comparisons between Rad53-FHA1 and Rad53-FHA2, with the former having a short helical insertion between the β2 and β3 strands (11, 12), whereas the latter has a much longer loop between the β5 and β6 strands (8). However, some structures determined so far show that the conserved residues located at the binding loops are positioned quite similarly in three-dimensional space. For example, despite a helical insertion in the loop between the β4 and β5 strands, the conserved Gly116, Arg117, Ser140, and His143 residues of the FHA domain of Chk2 can be spatially overlaid with the corresponding residues in Rad53-FHA1 (13). Moreover, regardless of the exact sequence, the conserved SXXH motif (S85NKH88 in FHA1) or its variant, such as S47RNQ50 in mouse polynucleotide kinase (PNK)-FHA (14), forms a protruding U-turn shape preceding the β5 strand.

In the mitotic checkpoint protein CHFR, the FHA domain forms a segment-swapped dimer, in which a C-terminal segment of one molecule occupies the position of the corresponding segment of the other molecule (15). It remains to be established whether this swapping has any biological implications, such as in protein oligomerization. However, human Chk2 and the Saccharomyces cerevisiae kinase Rad53 can self-dimerize and oligomerize, and their FHA domains are indispensable for these phosphorylation-dependent events (1618).

It is noteworthy that the structural topology of the FHA domain is similar to that of the MH2 (mothers against decapentaplegic homolog 2) domain of the tumor suppressor Smad (11, 19). However, although they have an identical core β-sandwich, the MH2 domain contains several helical insertions and has a different topology of its loops compared with that of the FHA domain (8, 11). Huse et al. proposed that the functions of these domains may also be conserved (20). A type I transforming growth factor–β (TGF-β) receptor–derived phosphopeptide interacts with MH2 in a region that coincides with the phosphopeptide-interaction region of FHA domains (20).

Combinatorial Library Approach and Ligand Specificity

The combinatorial peptide library approach was introduced in the early 1990s to study the ligand specificity of the Src homology 2 (SH2) domain (21) and has since been successfully applied to many protein-protein interaction domains, including the phosphotyrosine-binding (PTB) domain, WW, and 14-3-3 domains, which recognize short pThr-, pTyr-, or pSer-containing motifs (22). When the structure of the FHA domain was solved (8), its ligand specificity was not clear, despite some pioneering biological studies showing its phosphoprotein-binding role (2326). Three types of combinatorial libraries—pThr, pSer, and pTyr—with two or three residues on each side randomized (for example, XXXpTXXX) were thus used to elucidate ligand specificity. Two laboratories, using different screening and analytical methods, independently arrived at the conclusion that Rad53-FHA1 specifically recognizes pTXXD (11, 12). The specificity of Rad53-FHA2 was less certain, and both pYXL (27) and pTXXL (11, 12) have been suggested as potential recognition motifs. Overall, the results of library screening and biological studies of several FHA domains led to the suggestion of the “pThr+3 rule” for FHA domains, with the pThr and the pThr+3 residues being the primary and secondary recognition sites, respectively, and the preferred +3 residues falling into two major categories: Asp and Ile or Leu (11, 28). There are some exceptions. For example, the FHA domain of kinase-associated protein phosphatase (KAPP) prefers Ser or Ala (11, 29), whereas the RING-finger protein 8 (RNF8)–FHA domain displays a strong preference for Tyr and Phe at the pThr+3 position (30). The results from combinatorial library screenings are summarized in Table 2.

Table 2 Summary of ligand specificities derived from chemical (combinatorial library) and biological approaches.
View this table:

The structures of several complexes between FHA domains and phosphopeptides obtained from library screening have been reported (Table 1). An example is the crystal structure of human Chk2-FHA in complex with the synthetic peptide “HFD(pT)YLI,” which contains Ile at the pThr+3 position (Fig. 5, A and B) (13). In a separate study, the pT68XXL motif from Chk2 was also tested in vitro and shown, through NMR studies, to bind to Chk2-FHA with modest affinity (Kd of ~12.3 μM) (31).

Fig. 5

FHA-phosphopeptide complexes. (A) Crystal structure of the complex formed between Chk2-FHA and a pThr peptide [RHFD(pT)YLIRR] [Protein DataBank (PDB) ID: 1GXC] (13). The side chains (heavy atoms only) of three conserved FHA domain residues are shown in ball-and-stick representation, whereas the peptide residues (heavy atoms only) are in stick representation. The carbon, nitrogen, oxygen, and phosphorus atoms are shown in gray, cyan, red, and blue, respectively. (B) Surface charge distribution of the same complex. Positive, negative, and neutral potentials are represented in blue, red, and white, respectively. The peptide residues pThr and Ile (pThr+3) are highlighted in yellow; the remainder are in green. (C) Two-dimensional 1H-15N HSQC NMR spectra of Dun1-FHA in free (black) and complex (red) forms. The peptide is the doubly phosphorylated Rad53 SCD1 peptide 3NI(pT)QP(pT)QQST12. Several important residues that exhibit large chemical shift perturbations (for example, Arg62, Ser74, Lys129, and Ser130) or reappear in the complex form [Thr75, Arg102, and Asn103 (boxed)] are labeled. The top panel shows the changes in NHε of Arg.

The pThr-binding region in an FHA domain typically includes residues in the loops and turns between the following pairs of β strands: β3-β4, β4-β5, β6-β7, and β10-β11 (Fig. 5A). Overall, the bound peptide typically exists in an extended conformation around the pThr residue, and its binding induces little global conformational change in the FHA domain. The latter is also self-evident in two-dimensional 1H-15N heteronuclear single-quantum coherence (HSQC) experiments, which show relatively small changes in chemical shifts in the FHA domain upon peptide binding (10, 27, 3136), for example, Dun1 [DNA-damage un-inducible (Dun) mutant 1] (Fig. 5C). However, NMR analyses also indicated that there may be important changes in backbone dynamics, as evidenced by resonance broadening (34) and motion freezing (32, 33) of the residues at the binding site. A detailed dynamics study of the KAPP-FHA domain has revealed a net increase in backbone rigidity upon phosphopeptide binding (37).

The mode of pThr recognition is highly conserved (Fig. 6A). To facilitate the comparison of key pThr-binding residues in different FHA domains, we have listed the residue number and the loop (for example, β3-β4 stands for the loop between the two strands) of these residues of some of the FHA domains in Table 1. The phosphate group is anchored through salt bridges, hydrogen bonds, or both, by conserved Arg (Arg70 in FHA1) and Ser (Ser85 in FHA1) residues and nonconserved residues, such as the one immediately following the Ser; for example, Asn86 in FHA1, Lys46 in Ki67 (antigen identified by monoclonal antibody Ki-67) (38), Arg48 in mouse PNK, and Arg61 in RNF8 (11, 14, 32, 34) (Fig. 4A). In NMR studies, the amide proton of this conserved Ser residue in six different FHA domains—from Rad53 (FHA1 and 2), Chk2, Ki67, Dun1 (39), and nuclear inhibitor of protein phosphatase 1 (NIPP1) (8, 12, 31, 40)—shows a characteristic downfield shift to around 11.5 to 12.5 parts per million upon binding of the pThr-peptide (Fig. 5C), in further support of microstructural similarity at the binding site. The methyl group of the pThr residue makes contacts with a number of FHA residues, for example, Ser82, Arg83, Thr106, and Asn107 in FHA1 (11, 32), which could explain how the FHA domain is prevented from binding to pSer residues.

Fig. 6

Schematic representation of the structural basis of the pTXXD specificity of FHA1 for pTXXD compared with the specificity of FHA2 for pTXXL. (A) Illustration of the specificity of Thr versus Ser (highlighted by a number of interactions between pThr-γCH3 and the protein). Other interactions, demonstrated by intermolecular NOEs observed between conserved residues, nonconserved residues, or both, of the Rad53-FHA1 domain and the pThr peptide from Rad9, are also presented (32). The ionic interactions involving phosphate group, hydrophobic interactions, and the electrostatic interactions involving Arg83 are indicated in red, black, and green, respectively. (B) Illustration of specificity at the pThr and pThr+3 positions. Conserved and nonconserved residues among the two domains are highlighted in red and blue, respectively (41).

The specificity toward the pThr+3 residue is not conserved (Table 2). Whereas Rad53-FHA1 displays a clear preference for an Asp residue because of the presence of Arg83 in the β4-β5 loop, Rad53-FHA2 preferentially binds to Ile or Leu, even though it appears to have an Arg (Arg617) structurally equivalent to that of Rad53-FHA1. As revealed in the FHA2 structure, the latter is engaged in an intramolecular salt bridge with Asp683, which makes its guanidino group unavailable for intermolecular recognition (6, 27, 36). In light of this structural information, the binding specificities of these two domains can be altered by rational site-directed mutagenesis (41), as we have illustrated (Fig. 6B).

Although the library approach and the structural determination of FHA domain complexes with short phosphopeptides derived from consensus motifs have provided valuable information in early studies of FHA domains, these methods have had their shortcomings. First, the practical number of random sites is five, or six at most (for example, XXXpTXXX, where X is a randomized amino acid), which limits the ability to test the roles of residues remote from the pThr site. Second, the selection of random residues may be biased toward highly charged peptides because the electrostatic term in the free energy calculations of protein-ligand recognition seems to contribute more to binding affinity (42, 43). In addition, a large number of hydrophobic peptides may not dissolve in the binding buffer, thus reducing the number of possible candidate peptides that can be screened (44). Thus, many laboratories have actively pursued functional studies to identify binding partners and binding sites, and a wealth of insightful information has been generated.

Diversity of FHA-Ligand Interactions Based on Biological Approaches

Biological studies from many laboratories showed that the ligand specificity of FHA domains is much more diverse than that suggested by the so-called “pThr+3 rule” generalized from library screenings. It was found that, in addition to the pThrXX(Asp or Ile or Leu) specificity described above, there are three additional types of recognition specificities (Fig. 7).

Fig. 7

Illustration of four different mechanisms of ligand binding by FHA domains. (A) Schematic illustration of the “pThr+3” mechanism, in which some FHA domains recognize their targets mainly by the presence of the pThr and the residue at the +3 position (1113, 36). (B) The “N- and C-termini to pThr” mechanism of FHA domain interaction involves amino acid residues both N-terminal and C-terminal to the pThr residue (14). (C) The “pThr-pThr+3” mechanism was recently discovered for the yeast Dun1-FHA domain. Recognition specificity was determined by two pThr residues and the +3 position of the second pThr (33). (D) The “pThr + an extended binding surface” mechanism is required for the interaction between the Ki67-FHA domain and its binding partner NIFK (34).

FHA Domains with pThr+3 Specificities

FHA domains with specificities for pTXXD or pTXX(I/L/V).

The pTXXD or pTXX(I/L/V) specificities (Fig. 7A) identified by library screening have been verified in a number of biological studies (Table 2). For instance, S. cerevisiae Rad53-FHA1 and the Schizosaccharomyces pombe checking DNA synthesis protein 1 (Cds1)–FHA domain both prefer a highly charged Asp residue at the pThr+3 position (11, 45, 46). On the other hand, FHA2 from S. cerevisiae Rad53 and FHA domains from human Chk2 and mediator of DNA-damage checkpoint–1 (MDC1) [also known as nuclear factor with an amino-terminal FHA domain and a tandem repeat of BRCT (breast cancer susceptibility gene–1 C-terminal) domains (NFBD1)] prefer a nonpolar residue at the pThr+3 position (Leu, Ile, or Val) (11, 18, 47, 48). The PP2C-type phosphatases Ptc2 and Ptc3 are required in DNA checkpoint inactivation because of their interaction with Rad53, but only as a result of double-strand breaks induced by HO endonucleases (49). The interaction of Ptc2 or Ptc3 with Rad53 is mediated by the FHA1 domain (7-163) through a phosphorylation-dependent mechanism. Thr376 of Ptc2 (within a pTXXD motif) is the in vivo binding site for Rad53-FHA1 domain. Briefly, Ptc2 interacts with activated Rad53, dephosphorylates it, and consequently leads to cell cycle resumption and recovery after DNA damage (45).

Single FHA domain with multiple pThr+3 specificities.

Biological studies identified modifier of damage tolerance–1 (Mdt1) as a binding partner of Rad53-FHA1, with Thr305 as the recognition site (50). However, the +3 residue from Thr305 is a hydrophobic Ile residue and not a charged Asp. How can the same FHA1 domain bind to two different phosphopeptides, one containing pTXXD and the other containing pTXXI? A detailed comparison of the binding interactions between FHA1 and the pTXXI peptide from Mdt1 with those of FHA1 and the pTXXD peptide from Rad9 (51) is shown in Fig. 8. The FHA1-pTXXI complex does indeed retain the features of FHA1-pTXXD in recognizing the pThr moiety, which is anchored by hydrogen bonding or salt bridges to Arg70, Ser85, Asn86, and Thr106. However, although the Asp in the FHA1-pTXXD complex is involved in a strong charge-charge interaction with the Arg83 of FHA1, the Ile of the FHA1-pTXXI complex may contribute substantially to the binding affinity by mediating hydrophobic interactions with Gly135 and particularly Val136, which are located in the β10-β11 loop. Furthermore, residues other than pThr and the +3 residue also provide fine-tuning of the binding interactions between the FHA domain and its target peptide (51). In summary, structural data indicate that an FHA domain is able to bind to pTXXD and to pTXX(I/L/V) peptides by adapting to the different sequences of the phosphopeptide.

Fig. 8

Comparison of structures of FHA1 in complex with pTXXI and pTXXD phosphopeptides. (A) Surface-charge diagram of Rad53-FHA1 with a pThr peptide (SLEVpT192EADATFVQ) from Rad9 (PDB ID: 1J4Q) (32). (B) Surface-charge diagram of Rad53-FHA1 with a pThr peptide (NDPDpT305LEIYS) from Mdt1 (PDB ID: 2A0T) (51). (C) Stereoview showing the overlay of FHA1 structures in complex with pTXXI (PDB code 2A0T) and pTXXD (PDB code 1K3Q) phosphopeptides. The FHA1 domain is shown in a ribbon diagram colored in blue (FHA1-pTXXI complex) and green (FHA1-pTXXD complex). Several residues are highlighted: Val136 in FHA1-pTXXI, Arg83 in FHA1-pTXXD, and the peptide pThr and pThr+3 residues. The atoms are colored as for Fig. 5A. It appears that the pThr moiety is recognized in essentially the same way for both complexes. However, although there is strong ionic interaction between Asp(pT+3) and Arg83 in FHA1-pTXXD, the Ile(pThr+3) in FHA1-pTXXI contributes substantially to the binding through a hydrophobic interaction with the Val136 of FHA1 and so moves the residues C-terminal to pThr from the β6-β7 loop to the Val136-residing β10-β11 loop.

pThr+3 specificities identified from chemical versus biological approaches.

An example of a discrepancy between chemical and biological approaches was observed in Rad9, a suggested binding partner of Rad53-FHA1, which contains five possible ThrXXAsp motifs. Phosphopeptides corresponding to these five sites were synthesized and tested for their binding affinities for Rad53-FHA1. The peptide with the lowest Kd was S188LEV(pT)EADATFVQ200 (0.36 μM), which led to the suggestion that Thr192 of Rad9 was the likely binding site and the determination of the structure of the Rad9:Rad53-FHA1 complex (Fig. 8A) (32). However, in a subsequent report, the most favorable binding site of Rad53-FHA1 to Rad9 in vivo was shown to be Thr390, which has a Val at the pThr+3 position (48). The discrepancy between the two approaches regarding the exact pThr site of Rad9 serves as one caution about the biological relevance of the chemical library approach. However, there are many other cases where chemical and biological approaches identified the same ligand specificity (16, 18, 45, 49, 52, 53).

FHA Domains That Recognize Residues Both N-Terminal and C-Terminal to pThr

At least two FHA domains interact substantially with residues N-terminal to the pThr site (Fig. 7B). In murine PNK-FHA complexed with a peptide derived from its biological target XRCC4, only the residues N-terminal to pThr, cradled between the β2-β3 and β4-β5 loops, are essential for binding (14). This unusual binding behavior may be due to several basic residues in the FHA domain interacting with the acidic residues N-terminal to the pThr of the target protein (Fig. 9A). Analysis of a complex containing the FHA domain of Mycobacterium tuberculosis EmbR and a low-affinity phosphopeptide SLEV(pT)EADT (Kd = 185 μM) showed interactions similar to those observed in eukaryotic FHA-peptide complexes, with Asn348 interacting with Glu and Asp at positions pThr+1 and pThr+3, respectively. In addition, a nonconserved Leu residue in this bacterial FHA domain interacts with a Leu at the pThr-3 position of the peptide (54) (Fig. 9B). However, the biological relevance of these interactions remains to be validated because the phosphopeptide used in the structure was from S. cerevisiae Rad9 and not from the binding partner of EmbR in M. tuberculosis.

Fig. 9

Stereoview of structures of FHA domain complexes with phosphopeptide fragments derived from biological studies. (A) The binding of PNK-FHA to a pThr peptide YDES(pT)DEESEKK from XRCC4 (PDB ID: 1YJM) (14). (B) The binding of EmBR-FHA with a synthetic pThr peptide SLEV(pT)EADT (PDB ID: 2FF4) (54). (C) Structure of Rad53-FHA1 in complex with a singly phosphorylated Rad53-SCD1 peptide 3NI(pT)QPTQQST12 (PDB ID: 2JQI). (D) Structure of Dun1-FHA in complex with a doubly phosphorylated Rad53 SCD1 peptide 3NI(pT)QP(pT)QQST12 (PDB ID: 2JQL). Several important FHA residues are shown in ball-and-stick representation (heavy atoms and side chains only), whereas the peptide residues (heavy atoms only) are in stick representation. The atoms are colored similarly to those in Fig. 5A.

Dun1-FHA Contains Two pThr Sites

SCDs (SQ-TQ cluster domains) are abundant in various proteins, and there are two such domains in the Rad53 kinase of S. cerevisiae, which precede each of the two FHA domains. The Rad53-SCD1 domain consists of four Thr-Gln (TQ) sites at positions 5, 8, 12, and 15 (Fig. 4B) and is important for Rad53 dimerization and activation (55, 56). Furthermore, Stern and co-workers demonstrated that the Dun1-FHA domain interacts with the phosphorylated SCD1 of Rad53, which leads to activation of Dun1 (56). It is an intriguing question how such a phosphorylation cluster orchestrates the sequential activation of two separate kinases in a temporally ordered manner. Our own study indicates that Rad53-FHA1 recognizes singly phosphorylated SCD1, whereas Dun1-FHA specifically binds to SCD1 with dual phosphorylation at Thr5 and Thr8 (Fig. 9, C and D) (33). The structure of the dual pThr complex indicates that the first pThr-binding site is formed mainly by the conserved Arg60 residue and assisted by a nonconserved Arg62 located in the β3-β4 loop and that the second pThr-binding site consists of the mostly nonconserved Arg102 residue with possible assistance from Lys100 located in the β6-β7 loop. In addition, Ser11 at the +6 position relative to the first pThr also makes a considerable contribution to the binding, as evidenced by strong nuclear Overhauser effects (NOE) between this residue and Lys129 of Dun1-FHA in the β10-β11 loop. In vivo studies also showed that phosphorylation of both Thr5 and Thr8 on SCD1 is required for Dun1 to be fully activated and for the Dun1-dependent transcriptional DNA damage response, whereas single phosphorylation was sufficient for Rad53 to be fully activated and for Rad53-dependent cell survival. The existence of double phosphorylation of Thr5 and Thr8 and its enhancement upon DNA damage were further corroborated by mass spectrometry analyses (33).

The recognition of dual pThr residues by Dun1-FHA (Fig. 7C) is unprecedented among FHA domains. The results also shed light on the biological functions of the cluster of phosphorylation sites, suggesting a potential mechanism of sequential phosphorylation to direct a cascade of signaling events. Database searches have indicated that some other FHA domains could also possess dual pThr-recognizing specificities (33).

Ki67-FHA Recognizes an Extended Binding Surface

The Ki67 antigen protein, which contains 3256 amino acid residues, includes an FHA domain near its N terminus. Ki67 is widely used as an indicator of growth in cell populations because of its absence in resting cells (cells in G0 phase) and its readily detectable nuclear localization and association with condensed chromosomes during interphase and mitosis, respectively (57). The biological function of Ki67, however, is little known. In addition to its suggested role in cell cycle progression, its involvement in the organization of higher-order chromatin structures has also been postulated (58). Through yeast two-hybrid screening, Yoneda’s group identified human kinesin–like protein 2 (Hklp2) and human nucleolar protein interacting with the FHA domain of pKi67 (hNIFK) as the binding partners of the Ki67-FHA domain (59, 60) and proposed that Hklp2 is likely phosphorylated during the mitotic phase, during which its interaction with Ki67-FHA is at its strongest.

After unsuccessful attempts to identify short phosphopeptides that bind to Ki67-FHA, we showed that the synthetic fragment consisting of residues 226 to 269 of hNIFK binds tightly to Ki67, provided that Thr234 is phosphorylated. In vitro kinase assays showed that Thr234 is phosphorylated by glycogen synthase kinase 3 (GSK3), but only if Thr238 is first phosphorylated by cyclin-dependent kinase 1 (CDK1)–cyclin B (34, 40, 61). The structure of the Ki67-FHA complex with hNIFK (226-269) phosphorylated at Ser230, Thr234, and Thr238 was then solved by NMR spectroscopy, which showed an extended binding surface for protein-phosphoprotein interactions (Fig. 7D). A particularly interesting feature is that the β strand of the peptide stacks with the β sheet of Ki67-FHA (Fig. 10). In addition, the interaction of a large number of hydrophobic residues with the peptide was also considerably different from what was observed with other FHA domain complexes (61). This structure represents the most extensive structural information of a complex between an FHA domain and a phosphoprotein and clearly shows that the interaction goes beyond the short stretch surrounding the pThr site. This structure also provides a basis for the quantitative evaluation of specific interactions. For example, changing pThr234 to pSer234 led to a decrease in the binding affinity of Ki67-FHA for hNIFK (226-269) by a factor of 70 (Kd increased from 0.077 to 5.5 μM), whereas deleting the β strand from the peptide led to a decrease in the binding affinity by a factor of 180. On the other hand, individual substitutions of the –1, +1, +2, and +3 residues with Ala resulted in very small effects on binding affinity (a factor of <4), which is a clear indication that the “pThr+3 rule” does not apply to the Ki67-FHA domain. The residues surrounding pThr appear to regulate the phosphorylation of Thr234 by GSK3 (34).

Fig. 10

Stereoview of the Ki67-FHA:NIFK (226-269)3P complex (34). (A) The ribbon diagram shows overall interactions. Four FHA residues important for pThr234 binding together with the three phosphorylated residues are highlighted, and the two β strands engaged in intermolecular β sheets are numbered (PDB ID: 2AFF). (B) Detailed interactions in this complex involving pThr234 and the +1, +2, and +3 residues of hNIFK and the interface between the binding loops of Ki67-FHA and the α helix of hNIFK are shown. Carbon, nitrogen, oxygen, sulfur, and phosphorus atoms are colored in gray, cyan, red, yellow, and blue, respectively. Some of the predicted hydrogen bonds are depicted with dashed lines.

FHA Domains Mostly Use Loops for Ligand Binding and Specificity Control

It is beyond the scope of this review to make detailed comparisons between FHA domains and other protein-protein interaction domains. However, a cursory comparison of the FHA domain with three well-studied domains, 14-3-3 (specific for pSer and pThr) (62), SH2 (specific for pTyr) (63), and WW (specific for proline-rich sequences, or pThr-pSer containing trans-proline) (64) (Fig. 11), indicates that the FHA domain is unique in that its binding site mostly involves its loops. It is important to note that in addition to the β3-β4, β4-β5, and β6-β7 loops observed in early structural studies with synthetic peptides (11, 13, 32, 36), the involvement of the β10-β11 loop in pThr-peptide binding was clearly observed in later structural studies of Rad53-FHA1, Ki67-FHA, and Dun1-FHA complexed with their physiologically relevant targets. Briefly, intermolecular NOE assignments showed that Val136 in Rad53-FHA1 (51), Ile91 in Ki67-FHA (34), and Lys129 in Dun1-FHA (33) interacted with peptide residues C-terminal to the pThr. Furthermore, it was also suggested, on the basis of peptide-induced chemical shift perturbation experiments, that the β8-β9 loop of the FHA domain of kinase-associated protein phosphatase (KAPP), which apparently is the longest among the structurally determined FHA domains, could be involved in binding (10). Thus, because the loops and turns connecting β strands are typically flexible in conformation, variable in length, and divergent in sequence, particularly those (for example, the β10-β11 loop) beyond the core FHA homology region, this might explain the diverse ligand-binding specificities exhibited by various FHA domains, which in turn provides structural and functional versatility to FHA domain–containing proteins.

Fig. 11

Stereoviews of structures of other protein-protein interaction domains. (A) X-ray structure of the 14-3-3 domain in complex with a phosphopeptide RLYH(pS)LPA (PDB code: 1QJA) (134). (B) Solution structure of the C-terminal SH2 domain of PLC-γ1 in complex with a phosphopeptide DND(pY1021)IPLPDPK from the PDGF receptor (PDB code: 2PLD) (135). (C) X-ray structure of the pin1 WW domain with doubly phosphorylated serines followed by prolines T(pS)PT(pS)PS (although only the second pSer is involved in binding) from the C-terminal domain of RNA polymerase II (PDB code: 1F8A) (136). Phosphorylated residues (Tyr, Ser, or Thr) are shown in green, prolines in red, and the remainder of the peptide in purple.

Biological Functions of FHA Domains

The diverse biological functions of FHA domains are highlighted in Fig. 2. Here, we briefly review several FHA domain–containing proteins. Because it is not possible to cover all the published literature, we place more emphasis on the systems that have also been studied for their structures and ligand specificities.

Checkpoint Signaling in Humans: Chk2, RNF8, NBS1, and MDC1

Chk2.

Human Chk2 consists of an SCD and an FHA domain N-terminal to the kinase domain (Fig. 1B). Chk2, together with Chk1, plays a critical role in cell cycle regulation, DNA repair, and DNA damage–induced apoptosis at G1/S, S, and G2/M checkpoints (65) (Fig. 2). Chk2 relays the DNA damage signal from the serine-threonine kinases ATM (ataxia telangiectasia, mutated) or ATR (ATM and Rad3-related), which phosphorylate Chk2 at Thr68 in its SCD (66). Consequently, the Chk2-FHA domain binds to the pThr68 of a second Chk2 molecule, which leads to dimerization, oligomerization, or both of Chk2, followed by autophosphorylation at the kinase activation loop for complete activation of Chk2 (16, 18) (Fig. 12A). Activated Chk2 phosphorylates and interacts with its downstream effectors, which leads to either halted cell cycle progression to provide time to enable repair of DNA, or the triggering of apoptosis if damage cannot be repaired (65) (Fig. 2). Another upstream mediator, E3 identified by differential display (EDD), which is a human homolog of Drosophila melanogaster “hyperplastic discs” and functions as an E3 ubiquitin ligase, associates with Chk2 through its FHA domain. Knockdown of EDD by small interfering RNA (siRNA) inhibited DNA damage–dependent phosphorylation of Thr68, thus making Chk2 incapable of responding to DNA damage (67). Breast cancer susceptibility gene–1 (BRCA1), a substrate of Chk2, interacts with Chk2 through phosphorylation-dependent binding with its FHA domain (13).

Fig. 12

Illustration of FHA domain–mediated biological functions. (A) Diagram showing how the FHA domain of each Chk2 family member facilitates both transphosphorylation (that is, phosphorylation of Chk2 by another kinase) and autophosphorylations (that is, phosphorylation of Chk2 by itself), both of which are required for full activation of the kinase. (B) Diagram showing the assembly of the checkpoint proteins at the chromatin flanking DNA lesions in response to ionizing radiation (IR)–induced DNA double-strand breaks (DSBs). See main text for details.

Chk2 is a tumor suppressor; mutations in the Chk2 gene were identified in a subset family of malignancy-prone Li-Fraumeni syndromes and in sporadic human cancers (68). The Chk2-FHA functions as both a phosphorylation-dependent and -independent mediator to link upstream activators with downstream targets, according to studies of the oncogenic mutations (Arg117 → Gly117, Arg145 → Trp145, and Ile157 → Thr157) within the FHA domain (13, 16, 18, 69). The Arg117 → Gly117 mutation occurs in the conserved region for pThr-binding and attenuates the autokinase activity in response to DNA damage (69). The locations of the other two oncogenic mutations, Arg145 → Trp145 and Ile157 → Thr157, are remote from the binding pocket of pThr. It is interesting that the Arg145 → Trp145 mutant cannot bind to the original pThr ligand. Arg145 is located in the core structure (β5) of FHA β strands, and the Arg145 → Trp145 mutation destabilizes Chk2 and prevents the transphosphorylation activity of ATM (18, 70, 71). On the other hand, the Ile157 → Thr157 mutant behaves like the wild-type (WT) protein in terms of Chk2 activation; however, this mutant fails to bind to the downstream target of ectopically expressed BRCA1, p53, and cell division cycle 25A (CDC25A) (13, 72, 73).

RNF8.

A recent addition to the family of checkpoint regulators is RNF8 (Fig. 1A), a novel DNA damage–response protein that is recruited to the chromatin that flanks DNA lesions, where it colocalizes with phosphorylated H2AX, a variant of histone H2A, phosphorylated ATM, MDC1, BRCA1, 53BP1, RAP80, and NBS1 (30, 7476) (Fig. 12B). These interactions lead to the formation of ionizing radiation (IR)–induced foci (IRIF) in response to DNA double-strand breaks (30, 7476). Knockdown of RNF8 by siRNA prevents adequate cell cycle arrest at the G2/M checkpoint when IR is applied (30). RNF8 acts downstream of γH2AX (phosphorylated H2AX) and MDC1; the targeting of RNF8 to DNA lesions is facilitated by MDC1 through its phosphorylation-dependent interaction with the RNF8-FHA domain (30, 75, 76). Assembly of other downstream checkpoint proteins, such as BRCA1, 53BP1, and RAP80, at the IRIF requires RNF8 to ubiquitinate the histones flanking the site of the DNA lesion (30, 7476) (Fig. 12B). The RNF8-FHA domain shows a high specificity for Tyr and Phe at the pThr+3 position, as determined by the aforementioned library screening, and the structure of the complex it forms with phosphopeptide has been solved by x-ray crystallography (30). The RNF8-FHA–oriented motifs of pTXX(F/Y) can be narrowed down to the four-TQ cluster–containing sequence of MDC1 from residues 698 to 768; the consensus sequence for the four binding sites is TQXF, which is phosphorylated by ATM (30, 76). Kd values for the binding of the four synthetic phosphopeptides encompassing the binding regions of MDC1 to the RNF8-FHA domain are in the range of 3 to 11 μM (30). Deletion of this putative RNF8-FHA domain–binding region (residues 698 to 768) of MDC1 abolishes its interaction with RNF8 and prevents the assembly of downstream checkpoint proteins at sites of DNA lesions (30). Thus, the FHA domain is required for RNF8 functions in orchestrating the DNA damage response for the assembly of checkpoint proteins and signal transduction at the sites of DNA damage (30, 75, 76).

NBS1 and MDC1.

NBS1 and MDC1 (also known as NFBD1) are two other important checkpoint signaling proteins, both of which contain FHA and BRCT domains (Fig. 1A). NBS1 was first identified from a protein mutated in Nijmegen breakage syndrome, which is characterized by chromosome instability, radiosensitivity, and a high frequency of malignancies (77, 78). NBS1 also forms a complex with meiotic recombination 11 (MRE11) and RAD50, which presumably functions in sensing DNA double-strand breaks (DSBs) in the DNA damage–response pathway, as well as tethering and editing two DNA ends in the homologous recombination repair pathway (79) (Fig. 12B).

NBS1 is not only a downstream target of ATM, but also acts upstream to promote optimal ATM activation and its recruitment to the flanking region of DSBs (80, 81). The unwinding of DNA ends by the MRE11:RAD50:NBS1 (MRN) complex is essential for stimulating the monomerization and activation of ATM (82). Activated ATM then phosphorylates H2AX at Ser139 neighboring the DNA lesion that serves as a docking site (83), which in turn attracts the BRCT repeat of MDC1 through its pSer-binding activity (Fig. 12B) (84). Meanwhile, the consecutive SDTD (Ser-Asp-Thr-Asp) motifs of MDC1, when phosphorylated by casein kinase 2 (CK2), are bound by the FHA domain (85, 86) and the adjacent BRCT repeat (87, 88) of NBS1; such binding directs the accumulation of the MRN complex to damaged chromatin. Furthermore, the recruited MDC1 acts as a molecular scaffold by accommodating additional ATM molecules through interactions of its FHA domain with phosphorylated ATM (89). This therefore provides an important feedback loop that amplifies ATM-dependent γH2AX signals at sites of damage by which MDC1 may promote downstream effector proteins that accumulate after the induction of DSBs (89). Alternatively, the FHA domain of MDC1 also interacts directly with pThr68 of Chk2 (47) and with MRE11 (90), whereas knockdown of MDC1 by siRNA attenuates the phosphorylation of p53 at Ser20 by Chk2 and decreases IR-induced p53 stabilization, which results in a weakened apoptotic response (47). The knockdown also abates IRIF formation of the MRN complex, 53BP1, BRCA1, and γH2AX (91), which is suggestive of fundamental roles for MDC1 in mediating DNA-damage checkpoints.

PNK and ATPX in the Regulation of DNA Repair Pathways

In addition to DNA-damage signaling mediated by NBS1 and/or MDC1, which presumably facilitates or is essential for the repair of DNA DSBs by homologous recombination (9294), the FHA domain is also involved in other DNA-repair mechanisms. Mammalian PNK has DNA-kinase and DNA-phosphatase activities and is responsible for maintaining 5′-phosphate and 3′-hydroxyl termini at DNA-strand breaks, a prerequisite for polymerases and ligases operating in DNA repair (95). PNK is recruited to sites of base excision repair (BER) and nonhomologous end-joining (NHEJ) repair through its N-terminal FHA domain interacting with CK2-phosphorylated pThr residues of XRCC1 and XRCC4 (x-ray repair complementing defective repair in Chinese hamster cells 1 and 4, respectively), two key components of BER and NHEJ repair pathways that complex with DNA ligases III and IV, respectively (96, 97). In support of this, a crystallographical study of PNK shows that its FHA domain binds to an XRCC4-derived phosphopeptide (14); more specifically, it exhibits the highest selectivity for residues N-terminal of pThr, some selectivity for C-terminal residues, and no selectivity for the pThr+3 position (Fig. 9A) (96).

The FHA domain of aprataxin (APTX), a protein also involved in DNA repair and genomic stress (98), contains the same conserved residues involved in phosphopeptide recognition as does the PNK-FHA (14), which may suggest that the same ligand-sequence selectivity is shared between these two FHA domains. Indeed, APTX also forms complexes in vivo with XRCC1/ligase III and XRCC4/ligase IV and has been proposed to facilitate PNK activity during BER and NHEJ repair mechanisms (99, 100). Nevertheless, in contrast to the PNK-FHA, the APTX-FHA shows high specificity for a triply phosphorylated peptide derived from XRCC1 with a charged residue at the pThr+3 position (99). Taken together, these two FHA domains are likely to use different phosphopeptide-binding specificities to interact with the same binding partners in order to regulate each other’s activities. Recently, PALF [PNK and APTX-like FHA protein, also known as APLF (aprataxin and PNK-like factor)] was identified by two laboratories and shown to contain an FHA domain and a zinc finger–like CYR (Cys-Tyr-Arg) motif. This protein directly interacts with Ku86, ligase IV, and phosphorylated XRCC4 proteins; has endonuclease and exonuclease activities; and thus plays roles similar to this category of proteins in DNA damage (101, 102). The binding specificities of its FHA domain have not yet been determined.

Cds1, Rad53, and Dun1 in Checkpoint Signaling in Yeast

Here, we discuss Chk2 homologs in yeast: Cds1 in S. pombe and Rad53 and Dun1 in S. cerevisiae. The domain structures of these three yeast kinases are similar, but not identical, to Chk2 (Fig. 1B). Dun1 lacks an SCD preceding its FHA domain, whereas Rad53 contains two sets of SCD-FHA domains. The discussion below will focus on the roles played by the FHA domain in the activation of these kinases in response to DNA-damage or replication stress. As described above, Chk2 is activated by a two-stage process: FHA-independent phosphorylation by an upstream kinase, followed by FHA-dependent dimerization and autophosphorylation. The three yeast kinases largely follow the same two-stage processes; however, the molecular mechanisms mediated by their FHA domains vary substantially in each case (Fig. 12A).

Cds1.

In S. pombe Cds1, the FHA domain is involved in both phosphorylation of Cds1 by upstream kinases and autophosphorylation of Cds1 (46). Phosphorylation of Cds1 is mediated by the TQ repeats (Thr645 to Gln646 and Thr653 to Gln654, Fig. 4B) of mediator of replication checkpoint protein 1 (Mrc1) (46, 103). Upon replication stress, Cds1-FHA binds to pThr645 and pThr653 of Mrc1, which leads to the recruitment of Cds1 to the upstream kinase Rad3, which subsequently phosphorylates Thr11 of the Cds1-SCD (52, 53). The Cds1-FHA domain then binds to pThr11 of the Cds1-SCD, which leads to dimerization and autophosphorylation of Cds1 (46) (Fig. 12A). Note that both TQ repeats of Mrc1 share the consensus sequence surrounding pThr (Fig. 4B) and redundantly interact with Cds1-FHA to activate Cds1. Furthermore, the Asp residue in the pThr+3 position matches that identified in the phosphopeptide motif pTXXD in vitro, and mutation of both Asp residues to Ala (Asp648 → Ala648 and Asp658 → Ala658) results in hydroxyurea sensitivity (46). The Kd values of synthetic peptides encompassing pThr11 within Cds1-SCD and pThr653 within Mrc1-SCD for the Cds1-FHA domain are approximately 3.8 and 0.27 μM, respectively (46). In addition, Cds1 inhibits the activity of Mus81, a structure-specific endonuclease that plays a role in recombination, through its FHA domain. Thus, Cds1-FHA helps to maintain genomic integrity during certain types of replication stress (104).

Rad53.

Checkpoint responses of S. cerevisiae Rad53 are required for DNA damage–induced signaling and for cell cycle arrest. They are invoked in the event of DNA damage that occurs at various stages of the cell cycle (G1/S, S-phase progression, and G2/M transitions) or inhibition of DNA replication (105107). Rad53 is also essential for cell viability (108), normal cell growth, and transcriptional regulation (109). The mechanism of transactivation of Rad53 may be analogous to that of S. pombe Cds1 but could be substantially more complicated (Fig. 12A). Rad53 has two FHA domains, FHA1 (12) and FHA2 (8), which have become prototypical of studies of the structure-function relationship of the FHA domain since the landmark work by Stern’s group (24). These FHA domains may play diverse and overlapping roles in regulating Rad53 activation and in identifying and binding to prephosphorylated substrates (110, 111). In addition, Rad53 is recruited by two adaptor proteins, Rad9 and Mrc1 (a functional homolog of S. pombe Mrc1), to be phosphorylated by mitosis entry checkpoint protein 1 (Mec1) or telomere length–regulation protein 1(Tel1) in response to DNA-damage and replication stress (24, 112114). Involvement of FHA1 and FHA2 in the recruiting process remains to be firmly established, although hyperphosphorylated Rad9 preferentially binds to FHA2 (24), whereas both FHA1 and FHA2 are required for the Rad9-dependent activation of Rad53 (110, 111, 114, 115). Various SQ and TQ sequences from the Rad9-SCD (from residue 390 to 458, Fig. 4B) have redundant functions in their interactions with Rad53. The Kd values of synthetic phosphopeptides encompassing the first TQ sequence (pThr390) within the Rad9-SCD for FHA1 and FHA2 are 2.5 and 1.4 μM, respectively (48). Note that Rad9 and Mrc1 both interact with Rad53-FHA1 after the occurrence of DNA damage, as determined in a mass spectrometry–based proteome-wide study (116). Rad9 also functions redundantly with Mrc1, the major adaptor in replication checkpoint signaling (112, 113), to facilitate Rad53 activation in the Δmrc1 mutant upon hydroxyurea-induced replication stress (112).

When DNA damage occurs in vivo, S. cerevisiae Rad53 forms dimers or oligomers, which undergo autophosphorylation (117). Mutations in its FHA and SCD domains compromise its intermolecular interactions and autophosphorylation activity (55). However, detailed molecular mechanisms of the roles of FHA1 and FHA2 in the autophosphorylation of Rad53 remain to be established. Singly phosphorylated Rad53-SCD1 phosphopeptides interact with the FHA1, but not FHA2, domain of Rad53 in vitro, with phosphorylation at Thr5-Gln6 or Thr8-Gln9 motifs having the lowest Kd values (~10 to 20 μM) (33). Another study (118) showed that the topological order of FHA1 and FHA2 is critical to the autophosphorylation of Rad53 and the downstream transcriptional activation of RNR3 (ribonucleotide reductase 3), whereas this order does not affect phosphorylation of Rad53 by upstream kinases. A mutant Rad53 with swapped FHA domains has no kinase activity, although it can still be recruited to upstream kinases in response to DNA damage. This evidence implies that the intermolecular interaction between the FHA domain and the phosphorylated SCD in a specific spatial arrangement might be critical to the dimerization- or oligomerization-driven autophosphorylation of Rad53.

Evidence is accumulating that Rad53-FHA domains mediate not only the activation, but also the inactivation, of Rad53 during cell cycle checkpoints (45, 110, 111, 114, 115). A return to normal growth conditions after stress-induced responses is crucial for the cell, and this occurs as a result of recovery (after DNA repair) or adaptation (the resumption of proliferation despite limited irreparable damage) (45). The S. cerevisiae phosphatases Ptc2 and Ptc3 interact with Rad53 and are involved in checkpoint inactivation in response to DSBs induced by site-specific homothallic switching endonuclease (49). The interaction of Ptc2 and Ptc3 with Rad53 occurs through the FHA1 domain of Rad53 in a phosphorylation-dependent manner. The Kd value for binding of the synthetic phosphopeptides encompassing pThr376 of Ptc2 (pT376DAD) to Rad53-FHA1 is 2.3 μM (45). In vivo, this binding occurs when Ptc2 is phosphorylated by CK2 on Thr376, which leads to dephosphorylation of Rad53 and the resumption of the cell cycle after DNA damage (45).

Dun1.

Last, but not least, is the dramatic finding of an FHA domain with two specific pThr-binding sites (33) (Fig. 7C). S. cerevisiae Dun1 is a kinase that is phosphorylated and activated by Rad53 (119, 120). In vivo studies have indicated that the phosphorylated SCD1 of Rad53 interacts with Dun1-FHA (56) and leads Rad53 to phosphorylate Thr380 at the Dun1 activation loop, thus activating Dun1 (120) (Fig. 12A). Mutation of all four Thr residues to Ala in the TQ motifs of SCD1 abolishes the kinase activity of Rad53 and renders cells hypersensitive to DNA damage–causing reagents. The autokinase activity of Rad53 and Rad53-dependent survival in response to genotoxic stress can be restored by reverting any one of these mutated residues to Thr, whereas transduction of signals downstream of Dun1 remains retarded (33, 56). In vitro binding studies also indicate that the Rad53-FHA1 recognizes and binds to singly phosphorylated Rad53-SCD1 peptides. In contrast, the Dun1-FHA domain shows strongest binding to synthetic phosphopeptides of Rad53-SCD1 that contain simultaneously phosphorylated Thr5 and Thr8 residues (33). As discussed above, the results of structural and biological studies reveal a hierarchical regulation of the Rad53-Dun1 signaling cascade by a “phospho-counting mechanism” that involves both the Rad53-FHA1 and Dun1-FHA domains (33).

To date, whether and how Dun1-FHA regulates autophosphorylation of Dun1 remains elusive. A study indicated that mutations of Ser74 and His77 of the Dun1-FHA do not abolish its autokinase activity (119). Unlike human Chk2, S. cerevisiae Rad53, or S. pombe Cds1, the phosphorylation of Thr380 in the Dun1 activation loop, which activates the kinase activity of Dun1, occurs through its phosphorylation by Rad53 rather than through autophosphorylation (120, 121). This raises the question of why Dun1 still needs its autophosphorylation activity. However, it appears that autophosphorylation of Dun1 does occur and that it also plays a substantial role in regulating the kinase activity of Dun1 after DNA damage, because amino acid substitutions at the Dun1 autophosphorylation sites, Ser10 and Ser139, render cells sensitive to ultraviolet light and hydroxyurea treatment (120).

More than 10 Dun1-associated proteins have been identified, but the involvement of its FHA domain in these interactions is still unknown in most cases (122). Many of these proteins are related to the DNA damage response, but some are not. Reverse Spt phenotype 5 (Rsp5), an E3 ubiquitin ligase, is one of the latter, and it is considered an E3 ubiquitin ligase for suppressor of Mec lethality 1 (Sml1). Two other Dun1-associated proteins, translation machinery associated (TMA29, also known as Ymr226c) and phosphorylation inhibited by long-chain bases (PIL1, also known as Ygr086C), are induced by general cell stress, rather than by the occurrence of DNA damage. In addition, it has been suggested that Dun1 may be involved in a Rad53-independent checkpoint pathway (119). It is possible that future studies will uncover other mechanisms of the Dun1-FHA domain in these functions.

Roles for KAPP and DDL in Growth and Development in Plants

KAPP interacts with many receptor-like protein kinases (RLKs), through its kinase-interacting FHA domain in a phosphorylation-dependent manner (123). These RLKs, such as CLAVATA1 (CLV1), brassinosteroid-insensitive kinase 1 (BRI1), and BRI1-associated kinase 1 (BAK1), are associated with plant growth and development and are sensor proteins that trigger signaling cascades in response to changing environmental conditions. One study has shown that KAPP attenuates CLV1 activity in plant development (124). The interaction between KAPP and RLKs implicates FHA domains in signal transduction pathways of plants and, hence, in the regulation of plant development, environmental responses, and adaptation (29). Additionally, dawdle (ddl) plants with insertional mutations in the genes encoding other FHA domain–containing proteins of Arabidopsis thaliana show delayed development and defective roots, shoots, and flowers (125). The FHA domain–containing protein DAWDLE (DDL) plays a role in the biogenesis of microRNAs (miRNAs) and endogenous siRNAs (126). The FHA domain of DDL mediates its interaction with DICER-LIKE1 (DCL1), a protein that processes the conversion of pre-miRNAs to mature miRNAs in the nucleus (126). Smad nuclear interacting protein 1 (SNIP1), another FHA domain–containing protein and a human homolog of DDL, was also shown to perform a similar function, hence highlighting their conserved roles in miRNA biogenesis (126). The FHA domain–containing transcription activator NtFHA1 in Nicotiana tabacum, a likely functional analog of Fhl1 in yeast, was suggested to regulate cell growth and ribosomal RNA processing (127).

Role for EmbR in Bacterial Regulatory Pathways

Phosphorylation of proteins and enzymes by Ser-Thr protein kinases is a major means by which prokaryotic signaling mechanisms are regulated (128, 129). A comprehensive search and survey of bacterial FHA domains highlighted their possible roles in various processes, including regulation of cell shape, sporulation, transport, signal transduction, and ethambutol (Emb) resistance (130). Emb is a frontline antimycobacterial drug that targets the mycobacterial cell wall. A number of membrane-embedded arabinosyltransferases (encoded in the EmbCAB gene cluster) are involved in the synthesis of arabinogalactan and lipoarabinomannan, critical components of the mycobacterial cell wall (131). Expression of the genes encoding these arabinosyltransferases is positively regulated by the embR gene, which plays a role in resistance to Emb by modulating the activity level of the Emb-resistant arabinosyltransferase (132). Because deletion mutants of the stress-response Ser-Thr protein kinase PknH inhibit the expression of the emb genes, this indicates that bacterial cell wall synthesis and, hence, cell growth are controlled through a PknH-dependent pathway (133). The FHA domain–containing protein EmbR interacts with PknH in a phosphorylation-dependent manner and functions downstream of PknH, which implicates its FHA domain in the regulation of emb genes and thus in mycobacterial cell wall synthesis (54). EmbR has an N-terminal DNA binding domain, a bacterial transcriptional activation domain, and a C-terminal FHA domain, whose structure has been reported (54).

Future Perspectives

Despite the wealth of structural and functional information available, we are probably still at the early stage of a full understanding of the biological functions of FHA domains and of the structural bases of these functions. Even if a biological binding partner is clearly identified, the phosphorylation site(s) for binding by the FHA domain will need to be characterized. The next step will be to identify the kinase that phosphorylates this site. Subsequent preparation of the appropriate phosphoprotein for structural studies of protein-phosphoprotein complexes will also be a daunting task; the best that has been achieved so far is the preparation of a complex of Ki67-FHA and a phosphorylated 45–amino acid–residue fragment. Another important question is whether the phospho-counting signaling mechanism demonstrated in the Rad53-Dun1 cascade is shared in other signaling pathways involving FHA or other domains. Further advancement of these studies is a clear direction for FHA domain research in the next decade.

Acknowledgments

We thank former co-workers and collaborators who have contributed to the work in our laboratory, as well as the colleagues whose papers are cited in this review. We also wish to apologize to colleagues whose studies of the FHA domain may not have been highlighted here. Thanks also go to S. Hsiu-chien Chan for making Fig. 8, A and B. The work from our laboratory was supported by grants from the U.S. National Institutes of Health (CA69472 and CA87031) and the National Health Research Institute of Taiwan (EX95-9508NI).

References and Notes

View Abstract

Navigate This Article