The Pipsqueak Protein of Drosophila melanogasterBinds to GAGA Sequences through a Novel DNA-binding Domain*

Pipsqueak (Psq) belongs to a family of proteins defined by a phylogenetically old protein-protein interaction motif. Like the GAGA factor and other members of this family, Psq is an important developmental regulator in Drosophila, having pleiotropic functions during oogenesis, embryonic pattern formation, and adult development. The GAGA factor controls the transcriptional activation of homeotic genes and other genes by binding to control elements containing the GAGAG consensus motif. Binding is associated with formation of an open chromatin structure that makes the control regions accessible to transcriptional activators. We show here that Psq contains a novel DNA-binding domain, which binds, like the GAGA factor zinc finger DNA-binding domain, to target sites containing the GAGAG consensus motif. Binding is suppressed, as in the GAGA factor and other proteins of the family, by the associated protein-protein interaction motif. The DNA-binding domain, which we call the Psq domain, is identical with a previously identified region consisting of four tandem repeats of a conserved 50-amino acid sequence, the Psq motif. The Psq domain seems to be structurally related to known DNA-binding domains, both in its repetitive character and in the putative three-α-helix structure of the Psq motif, but it lacks the conserved sequence signatures of the classical eukaryotic DNA-binding motifs. Psq may thus represent the prototype of a new family of DNA-binding proteins.

Members of the BTB 1 /POZ protein family play important roles in development and reproduction of Drosophila melanogaster. These proteins contain a protein-protein interaction motif that was first identified in zinc finger proteins encoded by the Drosophila Broad-Complex and tramtrack genes (1,2), and later also in the bric à brac gene product (3). The domain, which was thereupon designated as BTB (Broad-Complex, Tramtrack, Bric à brac) domain (3), has since been found in proteins of a variety of species, as diverse as slime molds (4) and humans (for a review, see Ref. 5). Many of these proteins are DNA-binding C 2 H 2 zinc finger proteins, but the presence of the domain in a family of pox virus proteins (6) soon indicated that coupling to a DNA-binding domain is not mandatory. The domain is therefore also referred to as the POZ (pox virus, zinc finger) domain (7). In BTB/POZ proteins that contain a zinc finger DNA-binding motif, DNA binding is strongly inhibited by the BTB/POZ domain. This inhibitory effect on DNA binding is also observed in chimeric proteins in which the BTB/POZ domain is associated with a heterologous DNA-binding domain, for instance a POU domain (7). Inhibition of DNA binding appears to be the result of oligomerization through proteinprotein interactions mediated by the BTB/POZ domain.
The tendency of BTB/POZ proteins to oligomerize in solution and their localization in distinct nuclear substructures (7)(8)(9)(10) suggests that they might act by modifying chromatin structure (5). In fact, such a mode of action is supported by different lines of evidence for the E(var)3-93D product (11) and the GAGA factor of D. melanogaster. The two known isoforms of the GAGA factor, GAGA-519 and GAGA-581, which share the same BTB/POZ and zinc finger DNA-binding domains (12), bind to GAGAG consensus sites in the control regions of their target genes, which include homeotic genes (13) and heat shock genes (for a review, see Ref. 14). Binding leads to generation of an open chromatin conformation and thereby makes the control regions accessible to transcriptional activators (15)(16)(17). The role of the GAGA factor in establishing active chromatin structures is supported by the finding that a mutation of the GAGA factor-encoding Trithorax-like gene acts, like E(var)3-93D, as an enhancer of position effect variegation (18). Position effect variegation is observed as a mottled eye phenotype caused by the clonally inherited inactivation of a gene (e.g. white) that is juxtaposed to centromeric heterochromatin by a chromosomal rearrangement. Factors enhancing or suppressing position effect variegation are believed to be chromatin constituents or modulators of chromatin structure (reviewed in Ref. 19).
Pipsqueak (Psq) is a BTB/POZ protein with pleiotropic functions during development of D. melanogaster. Maternal psq function is required early in oogenesis (10,20), and psq is one of the posterior group genes that are responsible for pole cell formation and proper abdominal segmentation of the embryo (20). Moreover, psq directs correct localization of the gurken product, which is involved in establishment of the dorsoventral axis of the embryo (10). During metamorphosis, psq is required for formation of photoreceptors R3/R4 in the eye and for proper differentiation of other adult structures, such as wings and legs (21). The nuclear localization of Psq suggests that it acts, like many other BTB/POZ proteins, through binding to DNA (10). * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF084556 (honeybee psq cDNA AmPSQ). Access to the sequences of the prokaryotic recombinases is provided by Blocks Data base block BL00397 (see "Experimental Procedures"). The nucleotide sequence for D. melanogaster psq has been deposited in the EMBL data base under Accession Number X90986 (21) and in the GenBank TM data base under Gen-Bank TM accession number U48358 (10). The amino acid sequence of TKR can be accessed through the Swiss Protein Data base under Swiss-Prot number P14083 (22), and the nucleotide sequence of cosmid T01C1 has been deposited in the GenBank TM  However, none of the Psq isoforms that appear to be expressed by psq contains a homology to a known DNA-binding motif (10,21). Instead, a tandem repeat of four copies of a 50-amino acid sequence, called the Psq motif (10), is found at the C terminus. A sequence with homology to the Psq motif is present in another Drosophila protein, encoded by the Drosophila tyrosine kinase-related gene, Tkr (22), but the function of the motif remained unknown (10,21). Interestingly, cDNA structures suggest that, other than isoforms containing both the BTB/POZ domain and the Psq repeat region, there is at least one Psq isoform that contains the Psq repeats but lacks the BTB/POZ domain.
Here we show that the C-terminal region of Psq containing the four Psq repeats functions as a DNA-binding domain, which we call the Psq domain. The Psq domain binds to DNA consisting of direct repeats of the GAGAG consensus motif, which is otherwise recognized by the GAGA factor. Psq is thus the second GAGA-binding protein identified in D. melanogaster. Binding to DNA is sequence-specific and is inhibited by the BTB/POZ domain. The Psq domain appears to be structurally related to the DNA-binding domain of prokaryotic recombinases as well as to the eukaryotic homeodomain and Myb-like DNA-binding domain. However, it cannot be classified into any of these classes and may thus define a new class of eukaryotic DNA-binding domains.

EXPERIMENTAL PROCEDURES
Expression Library Screening-A cDNA library in -ZAPII (Stratagene) from brains of adult honeybees (a gift of D. Eisenhardt, Freie Universitä t Berlin) was screened essentially as described by Vinson et al. (23), except that the denaturation/renaturation steps were omitted. The sequence of oligonucleotide hspGAGA1, which was used for preparation of the concatenated probe, is given in Fig. 2E. A pBluescript phagemid containing cDNA AmPSQ (pBlueAmPSQ) was recovered by in vivo excision using helper phage R408 (Stratagene). AmPSQ was confirmed to originate from the honeybee by Southern hybridization and sequenced by automated cycle sequencing (MWG-BIOTECH).
Protein Synthesis and Mobility Shift DNA Binding Assay-Proteins were synthesized from the plasmid constructs in two parallel sets of reactions by cell-free in vitro transcription/translation (TNT T7/T3 Coupled Reticulocyte Lysate System; Promega) using radiolabeled or nonradiolabeled methionine. Labeled proteins were separated by SDS-PAGE and visualized by fluorography. Proteins of known molecular mass were separated in adjacent lanes for calibration of the gel.
Mobility shift DNA binding assays were performed as described (24), except that the concentration of EDTA was Յ1 mM, and 0.1 g of poly(dI-dC)⅐poly(dI-dC) was used. Binding reactions for the mobility shift shown in Isolation of a D. melanogaster Psq cDNA-A cDNA encoding the Psq domain of D. melanogaster was recovered from a 0 -4-h embryo cDNA library (a gift from E. Knust, Universität Dü sseldorf) by polymerase chain reaction using primers Dmpsq1 (5Ј-CAGCACGGTCTGGTCGA-CAGCGTTTGCGGCCCG -3Ј) and Dmpsq2 (5Ј-GCTGCCGGTGGTAC-CGCCCTAGGTGTTCAACAGC-3Ј). SalI and KpnI restriction sites, introduced into the polymerase chain reaction products through the primers, were cut with the appropriate enzymes, and the resulting fragments were cloned into an expression vector obtained by removing the SalI/KpnI insert from pBlueAmPSQ. The resulting plasmid thus encodes a 253-amino acid polypeptide, encompassing the D. melanogaster Psq domain, which is fused to the 37 N-terminal amino acids of Apis mellifera Psq.
Data Base Searches and Sequence Alignments-The "All non-redundant Protein Data base" (release May 10, 1998) was searched using the Advanced Blast2 search server at EMBL 3 employing standard parameters and substitution matrices from BLOSUM62 to BLOSUM30. The BLOCK maker program (25) 4 was used to create a 42-amino acid block of eight Psq repeat sequences. A search of this block against the Blocks Data base (release 10.1) (25) with LAMA (26) produced a single significant hit (expected chance occurrence Ͻ1%) to Block 397D, which contains the C-terminal part of the DNA-binding domains of prokaryotic recombinases. Secondary structure predictions for the Psq repeats performed by the Predator program at the EMBL (27) indicate a three-␣helix structure for all Psq repeats. Multiple alignments were calculated with ClustalX version 1.64b (28).

RESULTS AND DISCUSSION
Isolation of a cDNA Encoding the Psq Homologue of A. mellifera-To investigate the evolutionary conservation of proteins related to the GAGA factor of Drosophila melanogaster, we screened a cDNA expression library from honeybees (A. mellifera) using a radiolabeled probe containing GAGA factor binding sites. The probe was obtained by concatemerization of an oligonucleotide, hspGAGA1 (Fig. 2E), that had previously been used successfully to isolate GAGA factor cDNAs by expression screening (29). From 4.3 ϫ 10 6 cDNA clones screened, three clones reproducibly showed strong binding of the GAGA probe. Southern hybridization and sequencing revealed that only one of these clones represented a cDNA from A. mellifera. The cDNA, hereafter referred to as AmPSQ, contains an open reading frame encoding a 652-amino acid protein that shows high similarity to the Psq protein of D. melanogaster (Fig. 1). Although overall sequence conservation is only 33%, two domains show strikingly high conservation. First, the BTB/POZ domain is 73% identical with the BTB/POZ domain of D. melanogaster Psq. This means that despite the large evolutionary distance between D. melanogaster and A. mellifera, the degree of similarity between these domains is higher than the average identity of 53% between the BTB/POZ domains of different Drosophila proteins (30). Second, the region containing the four Psq repeats shows 80% identity with the Psq domain of the Drosophila protein. At the same time, not only the amino acid sequence of the single repeat units, but also the sequence of the repeats themselves is conserved (Figs. 1C and 5). We therefore conclude that AmPSQ encodes the Psq homologue of A. mellifera. The sequence between the BTB/POZ and Psq domains of D. melanogaster Psq contains several regions that are particularly rich in certain amino acid residues, including two glutaminerich regions and a region of 17 histidine residues alternating with other residues (Fig. 1A). Regions of this kind are believed to serve as interfaces for protein-protein interactions (31,32) and are frequently found in transcription factors (33,34). Poly-  Note that in all of the translation reactions truncated proteins are produced in addition to the full-length products. The sizes of the truncated products correspond with the predicted sizes of products that would result from the use of internal AUG triplets as start codons. These products thus represent N-terminal rather than C-terminal truncation derivatives that would result from precocious termination of translation or proteolytic cleavage. C, binding of full-length Psq and truncated Psq derivatives to DNA. Unlabeled translation products, produced by reactions run in parallel to the reactions analyzed in B, were incubated with radiolabeled hspGAGA2. The formation of protein-DNA complexes was then analyzed by a mobility shift assay. Complexes formed by full-length translation products are marked by arrows. Minor complexes marked by asterisks are probably formed by products derived from internal translation start sites (see above). The slowly migrating complex obtained with Psq ⌬384 may indicate an unusual conformation adopted by some Psq ⌬384-DNA complexes that is caused by the absence of 15 amino acid residues at the N terminus of the Psq domain. D, binding of full-length Psq and truncated Psq derivatives in the presence of magnesium. Binding of Psq ⌬240 and Psq ⌬333 is strongly stimulated by the addition of magnesium, compared with the binding of Psq ⌬384, which is not or is only marginally stimulated. Binding of full-length Psq is not observed also under these conditions, but binding of the putative N-terminal truncation products (see above) is enhanced. Amounts of protein loaded per lane and exposure time are reduced about 2-fold compared with the assay shown in C. The arrows and asterisks mark the same complexes as in C. E, structure of the oligonucleotides used for library screening (hspGAGA1) and as probes in mobility shift assays (hspGAGA2). glutamine tracts, in particular, have the potential to interact with components of the basal transcription machinery and can thus act as transcription activation domains (for a review, see Ref. 35). The interdomain sequence of the A. mellifera protein, which is only half as long as the interdomain sequence of Drosophila Psq, lacks these regions of special amino acid composition and contains a single asparagine-rich region instead. Since such asparagine-rich regions are also present in many transcription factors, particularly homeodomain proteins (34), we speculate that, despite their low sequence identity of only 13%, the interdomain regions of A. mellifera and D. melanogaster Psq have similar functions in both proteins.
The Psq Domain Is a Novel DNA-binding Domain-Since AmPSQ was isolated by an expression screen using a DNAbinding site probe, we reasoned that the conserved Psq domain might represent a novel DNA-binding domain. To test this hypothesis, we constructed plasmids encoding truncated Psq proteins ( Fig. 2A) and expressed these proteins, as well as full-length Psq, using a reticulocyte in vitro translation system. Fig. 2B shows that the Psq proteins were correctly expressed in the in vitro system. However, using a mobility shift DNA binding assay, we detected binding of neither full-length Psq nor of the truncated derivatives to the hspGAGA1 oligonucleotide (data not shown). Considering the repetitive structure of the Psq domain and the fact that a concatenated probe had been used for library screening, we suspected that a more extended binding sequence might be required for Psq binding. We therefore tested binding of the Psq proteins to oligonucleotide hsp-GAGA2, which is equivalent to the ligation product of two hspGAGA1 oligonucleotides (Fig. 2E). While binding of fulllength Psq could not be detected also with this probe (Fig. 2C,  lane 1), a truncated form of Psq that lacks an essential part of the BTB/POZ domain showed strong binding (Psq ⌬240, Fig.  2C, lane 2). This result is consistent with the finding of Bardwell and Treisman (7) that BTB/POZ domains generally inhibit the interaction of their associated DNA-binding domains with DNA. If the N-terminal truncation of Psq is further extended to essentially remove the N-terminal half of the protein, the DNA binding ability is retained (Psq ⌬333; Fig. 2C, lane 3). Even if the truncation advances into the first amino acid residues of the Psq domain (removing the first 15 amino acids of Psq repeat 1), the resulting 268 amino acid protein is still able to recognize the DNA target sequence (Psq ⌬384 ; Fig. 2C, lane 4). However, if the truncation removes the first two Psq repeats, DNA binding ability is lost (Psq ⌬474 ; Fig. 2C, lane 5). Truncations removing the Psq domain but leaving the BTB/POZ domain intact result in proteins that exert no specific DNA binding (Psq ⌬232 and Psq ⌬282; Fig. 2C, lanes 6 and 7). Since also full-length Psq does not appear to bind to DNA unless the BTB/POZ domain is removed, this result is not surprising. However, it is interesting to note that complexes formed by proteins that are probably translated from internal AUG start sites are absent in lanes loaded with the Psq ⌬232 and Psq ⌬282 translation products. Such complexes are observed in all lanes loaded with translation products of constructs encoding a Psq domain, whether they encode a BTB/POZ domain or not (Fig. 2C, lanes 1-4). This suggests that also products lacking both the Psq and BTB/POZ domain are unable to bind DNA. Taken together, these data indicate that the Psq domain is responsible for DNA binding of the Psq protein. Since Psq repeats 3 and 4 together are not sufficient for DNA binding (Psq ⌬474 ; Fig. 2C, lane 5), it seems likely that at least three complete repeat units are required for DNA binding. This is consistent with the requirement for a comparatively long DNA target sequence to detect binding. Interestingly, Horowitz and Berg (10) isolated a cDNA encoding a putative Psq isoform that resembles the Psq ⌬384 derivative in that it also lacks the first FIG. 3. The Psq domains of both A. mellifera and D. melanogaster Psq specifically recognize a DNA sequence containing the GAGAG-consensus motif. A, binding of A. mellifera Psq ⌬240 to radiolabeled hspGAGA2 was analyzed in the presence of increasing amounts of unlabeled competitor oligonucleotides. Oligonucleotides O3 (42 base pairs) and O30 (50 base pairs) uncover nonoverlapping sequences in the upstream region of the Drosophila Sgs-4 gene (see "Experimental Procedures"). Each of the indicated competitor DNAs was added in a 10-, 30-, and 100-fold molar excess over the probe. The faster migrating complex is formed by a truncated by-product of the translation reaction (see Fig. 2). B, binding of the D. melanogaster Psq domain to hspGAGA2, O3, and O30 was tested as in A.

FIG. 4. Psq and GAGA proteins bind independently to GAGA sequences in vitro.
In vitro translated full-length Psq and GAGA-519 proteins as well as the N-terminal truncation derivative Psq ⌬240 were incubated with radiolabeled hspGAGA2 either alone or in the indicated combinations. Protein-DNA complexes were then separated in a mobility shift gel. The products of each of the translation reactions form several complexes. Complexes marked by asterisks are likely to be formed by truncated products derived by the use of internal translation start sites (see Fig. 2). The arrow indicates the complex formed by the main Psq ⌬240 translation product. In vitro translated GAGA-519 forms a strong complex marked by an arrowhead. It is unknown if this complex is formed by full-length GAGA-519 or a derivative lacking the BTB/POZ domain. The GAGA-519 translation products also form a slowly migrating complex, marked by an open triangle, indicating the simultaneous binding of two or more protein molecules. Such oligomeric complexes are not observed with Psq ⌬240, suggesting that the length of hspGAGA2 is not sufficient for interaction with more than one Psq molecule. This is consistent with the putative structure and size of the Psq domain (see "Results and Discussion").
15 amino acid residues of Psq repeat 1. The deletions in both the D. melanogaster isoform and Psq ⌬384 leave Psq repeats 2-4 as well as the putative helix-turn-helix motif (see below) of Psq repeat 1 intact. Since Psq ⌬384 is still able to bind DNA, albeit with reduced affinity (see below), this is probably also true for the Drosophila isoform. It is intriguing to speculate that a reduced binding affinity or altered specificity of this isoform directs it to a specific subset of Psq binding sites in vivo.
DNA Binding Specificity of the Psq Domain-We next asked whether binding of the Psq domain to DNA depends on the nucleotide sequence of the DNA or if it recognizes DNA in a rather nonspecific manner. Fig. 3 shows that binding of Psq ⌬240 to the radiolabeled hspGAGA2 oligonucleotide is specifically inhibited already at a 10-fold molar excess of the nonlabeled oligonucleotide (Fig. 3A, lanes 2-4). Two different oligonucleotides, similar in length and G/C content to hspGAGA2 but of different nucleotide sequence (see "Experimental Procedures"), show no or only slight competition even at a 100-fold molar excess over the probe (Fig. 3A, lanes 5-7 and 8 -10). These data show that Psq binds to GAGA sequences with high sequence specificity.
Binding of Psq to DNA is greatly stimulated by the addition of magnesium to the incubation medium (Fig. 2D). Interestingly, this effect is observed with all DNA-binding Psq derivatives other than Psq ⌬384, which lacks the N-terminal 15 amino acid residues of Psq repeat 1. One possible explanation for this finding is that the deletion in Psq ⌬384 removes a Mg 2ϩ binding site, possibly located at the N terminus of the Psq domain, that is essential for high affinity binding. It is important to note that the deletion in Psq ⌬384 does not alter the binding specificity of the Psq domain, since binding of Psq ⌬384 shows the same sensitivity toward different competitor DNAs as binding of Psq ⌬240 (Fig. 3A and data not shown).
Since the Psq protein tested in these experiments was from A. mellifera, we wondered if the Psq domain of D. melanogaster would exhibit the same DNA binding specificity. We therefore cloned a 0.8-kilobase polymerase chain reaction fragment encoding the Drosophila Psq domain and expressed the polypeptide by in vitro-translation. When this polypeptide is incubated with the hspGAGA2 oligonucleotide, a strong complex is formed (Fig. 3B, lane 1). Formation of this complex is inhibited by increasing amounts of unlabeled hspGAGA2 but not by unrelated oligonucleotides shown before to be ineffective in competing for binding of the A. mellifera Psq domain (Fig. 3B,  lanes 2-10). Psq is thus the second GAGA-binding protein, other than the GAGA factor, that has been identified in D. melanogaster.
The similarity of the target sites recognized by Psq and GAGA factor suggests that binding of full-length Psq to GAGA sites in vivo might require the help of the GAGA factor. We therefore tested binding of full-length Psq and Psq ⌬240 to hspGAGA2 in the absence and presence of the full-length GAGA-519 isoform of D. melanogaster (Fig. 4). GAGA-519 in vitro translation products proved to be able to bind to hsp-GAGA2 with high affinity (Fig. 4, lane 2). As observed before, full-length Psq showed no binding and Psq ⌬240 showed strong binding to this oligonucleotide (Fig. 4, lanes 1 and 4). When either of these two proteins is mixed with the in vitro translated GAGA-519 isoform, the resulting pattern of DNA-protein complexes is the sum of the complex patterns observed in the presence of only the single proteins (Fig. 4, lanes 3 and 5). Thus, the GAGA-519 isoform does not seem to be able to promote DNA-binding of full-length Psq in vitro. It remains to be shown if Psq isoforms containing both the BTB/POZ and Psq domains in fact bind to GAGA sites or other DNA-binding sites in vivo or if they exert their functions independent of DNA binding. Since also isoforms lacking the BTB/POZ domain seem to be expressed in vivo (10), binding to GAGA sites or related target sites may be reserved to these isoforms.
The Psq Domain Is Similar to the DNA-binding Domain of Prokaryotic Recombinases-Psq cannot be easily assigned to any of the known families of eukaryotic DNA-binding proteins. Repeats with homology to the Psq motif are present in at least one additional Drosophila protein, the TKR protein (22), suggesting that also this protein is able to bind to DNA. Interestingly, Zollman et al. (30) identified a Drosophila BTB/POZ domain-encoding gene (BTB-III) that has an embryonic RNA distribution pattern very similar to that of Tkr and that maps to the same chromosomal position. It is thus interesting to speculate that the Tkr locus is more complex than previously supposed, encoding several protein isoforms, one of which con-  (38). Vertical arrows mark positions occupied by polar amino acid residues in all of the aligned sequences. Note that highest similarity is observed between A. mellifera and D. melanogaster Psq repeats, which take the same position within the Psq domain, indicating that not only are the repeats themselves conserved, but also the sequence of the repeats in the fly and honeybee Psq proteins. tains a BTB/POZ domain in addition to the Psq domain. Beyond Drosophila, a homology to the Psq motif is found in a polypeptide predicted by an open reading frame of Caenorhabditis elegans cosmid T01C1 (Ref. 10; Fig. 5). The Psq domain may thus define a new class of DNA-binding domains. At present, an extensive search of the protein sequence data bases reveals no other eukaryotic proteins with clear cut homology to the Psq motif. However, searching the Blocks Data base (25,26) with a multiple alignment of the eight Psq repeats reveals significant sequence similarities to the DNA-binding domain of prokaryotic recombinases (Fig. 5) and thereby provides a link between the Psq domain and the homeodomain, for which such similarities have been described as well (36). Cocrystal structures with DNA of two recombinases, Hin recombinase (37) and ␥␦-resolvase (38), show that their DNA-binding domains consist of three ␣-helices flanked by extended arms, which make contacts to the minor groove. The highest similarity between the Psq motif and the recombinase DNA-binding domain is observed within the C-terminal recognition helix, which forms a helix-turn-helix motif with helix 2 (Fig. 5) and inserts into the major groove. Remarkably, the recognition helix of members of the Hin recombinase family makes specific major groove contacts to a sequence that is clearly related to the GAGA motif (37). The Psq motif has the same size of about 50 amino acid residues as the recombinase DNA-binding domains, and secondary structure predictions for the Psq motif are compatible with the triple-helix structure of these domains. A similar triple-helix structure is formed by the homeodomain (39) and Myb DNA-binding domain (40). It is interesting to note that, like the Psq domain, also the Myb DNA-binding domain consists of imperfect tandem repeats of a conserved sequence motif (40). The Psq domain thus seems to be structurally related, both in its conformation and in its repetitive structure, to known DNA-binding motifs, but it eludes the classification into one of the prevalent categories of eukaryotic DNA-binding domains. Identification of additional members of the Psq family and determination of the structure of the Psq domain complexed with DNA will help to better define this new class of DNA-binding domains.